title
stringlengths 6
244
| abstract
stringlengths 19
5.09k
| label
stringclasses 10
values |
|---|---|---|
High-dimensional Varying Index Coefficient Models via Stein's Identity | We study the parameter estimation problem for a varying index coefficient
model in high dimensions. Unlike the most existing works that iteratively
estimate the parameters and link functions, based on the generalized Stein's
identity, we propose computationally efficient estimators for the
high-dimensional parameters without estimating the link functions. We consider
two different setups where we either estimate each sparse parameter vector
individually or estimate the parameters simultaneously as a sparse or low-rank
matrix. For all these cases, our estimators are shown to achieve optimal
statistical rates of convergence (up to logarithmic terms in the low-rank
setting). Moreover, throughout our analysis, we only require the covariate to
satisfy certain moment conditions, which is significantly weaker than the
Gaussian or elliptically symmetric assumptions that are commonly made in the
existing literature. Finally, we conduct extensive numerical experiments to
corroborate the theoretical results.
| stat |
Modeling Atmospheric Data and Identifying Dynamics: Temporal Data-Driven
Modeling of Air Pollutants | Atmospheric modelling has recently experienced a surge with the advent of
deep learning. Most of these models, however, predict concentrations of
pollutants following a data-driven approach in which the physical laws that
govern their behaviors and relationships remain hidden. With the aid of
real-world air quality data collected hourly in different stations throughout
Madrid, we present a case study using a series of data-driven techniques with
the following goals: (1) Find systems of ordinary differential equations that
model the concentration of pollutants and their changes over time; (2) assess
the performance and limitations of our models using stability analysis; (3)
reconstruct the time series of chemical pollutants not measured in certain
stations using delay coordinate embedding results.
| stat |
Ridge regression with adaptive additive rectangles and other piecewise
functional templates | We propose an $L_{2}$-based penalization algorithm for functional linear
regression models, where the coefficient function is shrunk towards a
data-driven shape template $\gamma$, which is constrained to belong to a class
of piecewise functions by restricting its basis expansion. In particular, we
focus on the case where $\gamma$ can be expressed as a sum of $q$ rectangles
that are adaptively positioned with respect to the regression error. As the
problem of finding the optimal knot placement of a piecewise function is
nonconvex, the proposed parametrization allows to reduce the number of
variables in the global optimization scheme, resulting in a fitting algorithm
that alternates between approximating a suitable template and solving a convex
ridge-like problem. The predictive power and interpretability of our method is
shown on multiple simulations and two real world case studies.
| stat |
Amazon Forest Fires Between 2001 and 2006 and Birth Weight in Porto
Velho | Birth weight data (22,012 live-births) from a public hospital in Porto Velho
(Amazon) was used in multiple statistical models to assess the effects of
forest-fire smoke on human reproductive outcome. Mean birth weights for girls
(3,139 g) and boys (3,393 g) were considered statistically different (p-value <
2.2e-16). Among all models analyzed, the means were considered statistically
different only when treated as a function of month and year (p-value = 0.0989,
girls and 0.0079, boys) . The R 2 statistics indicate that the regression
models considered are able to explain 65 % (girls) and 54 % (boys) of the
variation of the mean birth weight.
| stat |
A Unified Framework for Causal Inference with Multiple Imputation Using
Martingale | Multiple imputation is widely used to handle confounders missing at random in
causal inference. Although Rubin's combining rule is simple, it is not clear
whether or not the standard multiple imputation inference is consistent when
coupled with the commonly-used average causal effect (ACE) estimators. This
article establishes a unified martingale representation for the average causal
effect (ACE) estimators after multiple imputation. This representation invokes
the wild bootstrap inference to provide consistent variance estimation. Our
framework applies to asymptotically normal ACE estimators, including the
regression imputation, weighting, and matching estimators. We extend to the
scenarios when both outcome and confounders are subject to missingness and when
the data are missing not at random.
| stat |
Regression for Copula-linked Compound Distributions with Applications in
Modeling Aggregate Insurance Claims | In actuarial research, a task of particular interest and importance is to
predict the loss cost for individual risks so that informative decisions are
made in various insurance operations such as underwriting, ratemaking, and
capital management. The loss cost is typically viewed to follow a compound
distribution where the summation of the severity variables is stopped by the
frequency variable. A challenging issue in modeling such outcome is to
accommodate the potential dependence between the number of claims and the size
of each individual claim. In this article, we introduce a novel regression
framework for compound distributions that uses a copula to accommodate the
association between the frequency and the severity variables, and thus allows
for arbitrary dependence between the two components. We further show that the
new model is very flexible and is easily modified to account for incomplete
data due to censoring or truncation. The flexibility of the proposed model is
illustrated using both simulated and real data sets. In the analysis of
granular claims data from property insurance, we find substantive negative
relationship between the number and the size of insurance claims. In addition,
we demonstrate that ignoring the frequency-severity association could lead to
biased decision-making in insurance operations.
| stat |
Multi-sample Estimation of Bacterial Composition Matrix in Metagenomics
Data | Metagenomics sequencing is routinely applied to quantify bacterial abundances
in microbiome studies, where the bacterial composition is estimated based on
the sequencing read counts. Due to limited sequencing depth and DNA dropouts,
many rare bacterial taxa might not be captured in the final sequencing reads,
which results in many zero counts. Naive composition estimation using count
normalization leads to many zero proportions, which tend to result in
inaccurate estimates of bacterial abundance and diversity. This paper takes a
multi-sample approach to the estimation of bacterial abundances in order to
borrow information across samples and across species. Empirical results from
real data sets suggest that the composition matrix over multiple samples is
approximately low rank, which motivates a regularized maximum likelihood
estimation with a nuclear norm penalty. An efficient optimization algorithm
using the generalized accelerated proximal gradient and Euclidean projection
onto simplex space is developed. The theoretical upper bounds and the minimax
lower bounds of the estimation errors, measured by the Kullback-Leibler
divergence and the Frobenius norm, are established. Simulation studies
demonstrate that the proposed estimator outperforms the naive estimators. The
method is applied to an analysis of a human gut microbiome dataset.
| stat |
Exploiting offshore wind and solar resources in the Mediterranean using
ERA5 reanalysis data | Commercial electricity production from marine renewable sources is becoming a
necessity at a global scale. Offshore wind and solar resources can be combined
to reduce construction and maintenance costs. In this respect, the aim of this
study is two-fold: i) analyse offshore wind and solar resource and their
variability in the Mediterranean Sea at the annual and seasonal scales based on
the recently published ERA5 reanalysis dataset, and; ii) perform a preliminary
assessment of some important features of complementarity, synergy, and
availability of the examined resources using an event-based probabilistic
approach. A robust coefficient of variation is introduced to examine the
variability of each resource and a joint coefficient of variation is
implemented for the first time to evaluate the joint variability of offshore
wind and solar potential. The association between the resources is examined by
introducing a robust measure of correlation, along with the Pearson's r and
Kendall's tau correlation coefficient and the corresponding results are
compared. Several metrics are used to examine the degree of complementarity
affected by variability and intermittency issues. Areas with high potential and
low variability for both resources include the Aegean and Alboran seas, while
significant synergy (over 52%) is identified in the gulfs of Lion, Gabes and
Sidra, Aegean Sea and northern Cyprus Isl. The advantage of combining these two
resources is highlighted at selected locations in terms of the monthly energy
production.
| stat |
Bayesian Variable Selection for Non-Gaussian Responses: A Marginally
Calibrated Copula Approach | We propose a new highly flexible and tractable Bayesian approach to undertake
variable selection in non-Gaussian regression models. It uses a copula
decomposition for the joint distribution of observations on the dependent
variable. This allows the marginal distribution of the dependent variable to be
calibrated accurately using a nonparametric or other estimator. The family of
copulas employed are `implicit copulas' that are constructed from existing
hierarchical Bayesian models widely used for variable selection, and we
establish some of their properties. Even though the copulas are
high-dimensional, they can be estimated efficiently and quickly using Markov
chain Monte Carlo (MCMC). A simulation study shows that when the responses are
non-Gaussian the approach selects variables more accurately than contemporary
benchmarks. A real data example in the Web Appendix illustrates that accounting
for even mild deviations from normality can lead to a substantial increase in
accuracy. To illustrate the full potential of our approach we extend it to
spatial variable selection for fMRI. Using real data, we show our method allows
for voxel-specific marginal calibration of the magnetic resonance signal at
over 6,000 voxels, leading to an increase in the quality of the activation
maps.
| stat |
A Change-Point Based Control Chart for Detecting Sparse Changes in
High-Dimensional Heteroscedastic Data | Because of the curse-of-dimensionality, high-dimensional processes present
challenges to traditional multivariate statistical process monitoring (SPM)
techniques. In addition, the unknown underlying distribution and complicated
dependency among variables such as heteroscedasticity increase uncertainty of
estimated parameters, and decrease the effectiveness of control charts. In
addition, the requirement of sufficient reference samples limits the
application of traditional charts in high dimension low sample size scenarios
(small n, large p). More difficulties appear in detecting and diagnosing
abnormal behaviors that are caused by a small set of variables, i.e., sparse
changes. In this article, we propose a changepoint based control chart to
detect sparse shifts in the mean vector of high-dimensional heteroscedastic
processes. Our proposed method can start monitoring when the number of
observations is a lot smaller than the dimensionality. The simulation results
show its robustness to nonnormality and heteroscedasticity. A real data example
is used to illustrate the effectiveness of the proposed control chart in
high-dimensional applications. Supplementary material and code are provided
online.
| stat |
Envelope Methods with Ignorable Missing Data | Envelope method was recently proposed as a method to reduce the dimension of
responses in multivariate regressions. However, when there exists missing data,
the envelope method using the complete case observations may lead to biased and
inefficient results. In this paper, we generalize the envelope estimation when
the predictors and/or the responses are missing at random. Specifically, we
incorporate the envelope structure in the expectation-maximization (EM)
algorithm. As the parameters under the envelope method are not pointwise
identifiable, the EM algorithm for the envelope method was not straightforward
and requires a special decomposition. Our method is guaranteed to be more
efficient, or at least as efficient as, the standard EM algorithm. Moreover,
our method has the potential to outperform the full data MLE. We give
asymptotic properties of our method under both normal and non-normal cases. The
efficiency gain over the standard EM is confirmed in simulation studies and in
an application to the Chronic Renal Insufficiency Cohort (CRIC) study.
| stat |
Incorporating Dynamicity of Transportation Network with Multi-Weight
Traffic Graph Convolution for Traffic Forecasting | Traffic forecasting problem remains a challenging task in the intelligent
transportation system due to its spatio-temporal complexity. Although temporal
dependency has been well studied and discussed, spatial dependency is
relatively less explored due to its large variations, especially in the urban
environment. In this study, a novel graph convolutional network model,
Multi-Weight Traffic Graph Convolutional (MW-TGC) network, is proposed and
applied to two urban networks with contrasting geometric constraints. The model
conducts graph convolution operations on speed data with multi-weighted
adjacency matrices to combine the features, including speed limit, distance,
and angle. The spatially isolated dimension reduction operation is conducted on
the combined features to learn the dependencies among the features and reduce
the size of the output to a computationally feasible level. The output of
multi-weight graph convolution is applied to the sequence-to-sequence model
with Long Short-Term Memory units to learn temporal dependencies. When applied
to two urban sites, urban-core and urban-mix, MW-TGC network not only
outperformed the comparative models in both sites but also reduced variance in
the heterogeneous urban-mix network. We conclude that MW-TGC network can
provide a robust traffic forecasting performance across the variations in
spatial complexity, which can be a strong advantage in urban traffic
forecasting.
| stat |
A Model of a Randomized Experiment with an Application to the PROWESS
Clinical Trial | I develop a model of a randomized experiment with a binary intervention and a
binary outcome. Potential outcomes in the intervention and control groups give
rise to four types of participants. Fixing ideas such that the outcome is
mortality, some participants would live regardless, others would be saved,
others would be killed, and others would die regardless. These potential
outcome types are not observable. However, I use the model to develop
estimators of the number of participants of each type. The model relies on the
randomization within the experiment and on deductive reasoning. I apply the
model to an important clinical trial, the PROWESS trial, and I perform a Monte
Carlo simulation calibrated to estimates from the trial. The reduced form from
the trial shows a reduction in mortality, which provided a rationale for FDA
approval. However, I find that the intervention killed two participants for
every three it saved.
| stat |
A flexible forecasting model for production systems | This paper discusses desirable properties of forecasting models in production
systems. It then develops a family of models which are designed to satisfy
these properties: highly customizable to capture complex patterns; accommodates
a large variety of objectives; has interpretable components; produces robust
results; has automatic changepoint detection for trend and seasonality; and
runs fast -- making it a good choice for reliable and scalable production
systems. The model allows for seasonality at various time scales,
events/holidays, and change points in trend and seasonality. The volatility is
fitted separately to maintain flexibility and speed and is allowed to be a
function of specified features.
| stat |
Linear Optimal Transport Embedding: Provable fast Wasserstein distance
computation and classification for nonlinear problems | Discriminating between distributions is an important problem in a number of
scientific fields. This motivated the introduction of Linear Optimal
Transportation (LOT), which embeds the space of distributions into an
$L^2$-space. The transform is defined by computing the optimal transport of
each distribution to a fixed reference distribution, and has a number of
benefits when it comes to speed of computation and to determining
classification boundaries. In this paper, we characterize a number of settings
in which LOT embeds families of distributions into a space in which they are
linearly separable. This is true in arbitrary dimension, and for families of
distributions generated through perturbations of shifts and scalings of a fixed
distribution. We also prove conditions under which the $L^2$ distance of the
LOT embedding between two distributions in arbitrary dimension is nearly
isometric to Wasserstein-2 distance between those distributions. This is of
significant computational benefit, as one must only compute $N$ optimal
transport maps to define the $N^2$ pairwise distances between $N$
distributions. We demonstrate the benefits of LOT on a number of distribution
classification problems.
| stat |
Neural Group Actions | We introduce an algorithm for designing Neural Group Actions, collections of
deep neural network architectures which model symmetric transformations
satisfying the laws of a given finite group. This generalizes involutive neural
networks $\mathcal{N}$, which satisfy $\mathcal{N}(\mathcal{N}(x))=x$ for any
data $x$, the group law of $\mathbb{Z}_2$. We show how to optionally enforce an
additional constraint that the group action be volume-preserving. We
conjecture, by analogy to a universality result for involutive neural networks,
that generative models built from Neural Group Actions are universal
approximators for collections of probabilistic transitions adhering to the
group laws. We demonstrate experimentally that a Neural Group Action for the
quaternion group $Q_8$ can learn how a set of nonuniversal quantum gates
satisfying the $Q_8$ group laws act on single qubit quantum states.
| stat |
On Bayesian inference for the Extended Plackett-Luce model | The analysis of rank ordered data has a long history in the statistical
literature across a diverse range of applications. In this paper we consider
the Extended Plackett-Luce model that induces a flexible (discrete)
distribution over permutations. The parameter space of this distribution is a
combination of potentially high-dimensional discrete and continuous components
and this presents challenges for parameter interpretability and also posterior
computation. Particular emphasis is placed on the interpretation of the
parameters in terms of observable quantities and we propose a general framework
for preserving the mode of the prior predictive distribution. Posterior
sampling is achieved using an effective simulation based approach that does not
require imposing restrictions on the parameter space. Working in the Bayesian
framework permits a natural representation of the posterior predictive
distribution and we draw on this distribution to address the rank aggregation
problem and also to identify potential lack of model fit. The flexibility of
the Extended Plackett-Luce model along with the effectiveness of the proposed
sampling scheme are demonstrated using several simulation studies and real data
examples.
| stat |
Model-based inference of conditional extreme value distributions with
hydrological applications | Multivariate extreme value models are used to estimate joint risk in a number
of applications, with a particular focus on environmental fields ranging from
climatology and hydrology to oceanography and seismic hazards. The
semi-parametric conditional extreme value model of Heffernan and Tawn (2004)
involving a multivariate regression provides the most suitable of current
statistical models in terms of its flexibility to handle a range of extremal
dependence classes. However, the standard inference for the joint distribution
of the residuals of this model suffers from the curse of dimensionality since
in a $d$-dimensional application it involves a $d-1$-dimensional non-parametric
density estimator, which requires, for accuracy, a number points and
commensurate effort that is exponential in $d$. Furthermore, it does not allow
for any partially missing observations to be included and a previous proposal
to address this is extremely computationally intensive, making its use
prohibitive if the proportion of missing data is non-trivial. We propose to
replace the $d-1$-dimensional non-parametric density estimator with a
model-based copula with univariate marginal densities estimated using kernel
methods. This approach provides statistically and computationally efficient
estimates whatever the dimension, $d$ or the degree of missing data. Evidence
is presented to show that the benefits of this approach substantially outweigh
potential mis-specification errors. The methods are illustrated through the
analysis of UK river flow data at a network of 46 sites and assessing the
rarity of the 2015 floods in north west England.
| stat |
Time-varying Autoregression with Low Rank Tensors | We present a windowed technique to learn parsimonious time-varying
autoregressive models from multivariate timeseries. This unsupervised method
uncovers interpretable spatiotemporal structure in data via non-smooth and
non-convex optimization. In each time window, we assume the data follow a
linear model parameterized by a system matrix, and we model this stack of
potentially different system matrices as a low rank tensor. Because of its
structure, the model is scalable to high-dimensional data and can easily
incorporate priors such as smoothness over time. We find the components of the
tensor using alternating minimization and prove that any stationary point of
this algorithm is a local minimum. We demonstrate on a synthetic example that
our method identifies the true rank of a switching linear system in the
presence of noise. We illustrate our model's utility and superior scalability
over extant methods when applied to several synthetic and real-world example:
two types of time-varying linear systems, worm behavior, sea surface
temperature, and monkey brain datasets.
| stat |
Inference in experiments conditional on observed imbalances in
covariates | Double blind randomized controlled trials are traditionally seen as the gold
standard for causal inferences as the difference-in-means estimator is an
unbiased estimator of the average treatment effect in the experiment. The fact
that this estimator is unbiased over all possible randomizations does not,
however, mean that any given estimate is close to the true treatment effect.
Similarly, while pre-determined covariates will be balanced between treatment
and control groups on average, large imbalances may be observed in a given
experiment and the researcher may therefore want to condition on such
covariates using linear regression. This paper studies the theoretical
properties of both the difference-in-means and OLS estimators conditional on
observed differences in covariates. By deriving the statistical properties of
the conditional estimators, we can establish guidance for how to deal with
covariate imbalances. We study both inference with OLS, as well as with a new
version of Fisher's exact test, where the randomization distribution comes from
a small subset of all possible assignment vectors.
| stat |
Prior-preconditioned conjugate gradient method for accelerated Gibbs
sampling in "large $n$ & large $p$" sparse Bayesian regression | In a modern observational study based on healthcare databases, the number of
observations and of predictors typically range in the order of $10^5$ ~ $10^6$
and of $10^4$ ~ $10^5$. Despite the large sample size, data rarely provide
sufficient information to reliably estimate such a large number of parameters.
Sparse regression techniques provide potential solutions, one notable approach
being the Bayesian methods based on shrinkage priors. In the "large $n$ & large
$p$" setting, however, posterior computation encounters a major bottleneck at
repeated sampling from a high-dimensional Gaussian distribution, whose
precision matrix $\Phi$ is expensive to compute and factorize. In this article,
we present a novel algorithm to speed up this bottleneck based on the following
observation: we can cheaply generate a random vector $b$ such that the solution
to the linear system $\Phi \beta = b$ has the desired Gaussian distribution. We
can then solve the linear system by the conjugate gradient (CG) algorithm
through matrix-vector multiplications by $\Phi$, without ever explicitly
inverting $\Phi$. Rapid convergence of CG in this specific context is achieved
by the theory of prior-preconditioning we develop. We apply our algorithm to a
clinically relevant large-scale observational study with $n$ = 72,489 patients
and $p$ = 22,175 clinical covariates, designed to assess the relative risk of
adverse events from two alternative blood anti-coagulants. Our algorithm
demonstrates an order of magnitude speed-up in the posterior computation.
| stat |
Quantum-inspired Machine Learning on high-energy physics data | One of the most challenging big data problems in high energy physics is the
analysis and classification of the data produced by the Large Hadron Collider
at CERN. Recently, machine learning techniques have been employed to tackle
such challenges, which, despite being very effective, rely on classification
schemes that are hard to interpret. Here, we introduce and apply a
quantum-inspired machine learning technique and, exploiting tree tensor
networks, we show how to efficiently classify b-jet events in proton-proton
collisions at LHCb and to interpret the classification results. In particular,
we show how to select important features and adapt the network geometry based
on information acquired in the learning process. Moreover, the tree tensor
network can be adapted for optimal precision or fast response in time without
the need for repeating the learning process. This paves the way to
high-frequency real-time applications as needed for current and future LHC
event classification to trigger events at the tens of MHz scale.
| stat |
A Maximin $\Phi_{p}$-Efficient Design for Multivariate GLM | Experimental designs for a generalized linear model (GLM) often depend on the
specification of the model, including the link function, the predictors, and
unknown parameters, such as the regression coefficients. To deal with
uncertainties of these model specifications, it is important to construct
optimal designs with high efficiency under such uncertainties. Existing methods
such as Bayesian experimental designs often use prior distributions of model
specifications to incorporate model uncertainties into the design criterion.
Alternatively, one can obtain the design by optimizing the worst-case design
efficiency with respect to uncertainties of model specifications. In this work,
we propose a new Maximin $\Phi_p$-Efficient (or Mm-$\Phi_p$ for short) design
which aims at maximizing the minimum $\Phi_p$-efficiency under model
uncertainties. Based on the theoretical properties of the proposed criterion,
we develop an efficient algorithm with sound convergence properties to
construct the Mm-$\Phi_p$ design. The performance of the proposed Mm-$\Phi_p$
design is assessed through several numerical examples.
| stat |
Improved Inference of Gaussian Mixture Copula Model for Clustering and
Reproducibility Analysis using Automatic Differentiation | Copulas provide a modular parameterization of multivariate distributions that
decouples the modeling of marginals from the dependencies between them.
Gaussian Mixture Copula Model (GMCM) is a highly flexible copula that can model
many kinds of multi-modal dependencies, as well as asymmetric and tail
dependencies. They have been effectively used in clustering non-Gaussian data
and in Reproducibility Analysis, a meta-analysis method designed to verify the
reliability and consistency of multiple high-throughput experiments. Parameter
estimation for GMCM is challenging due to its intractable likelihood. The best
previous methods have maximized a proxy-likelihood through a Pseudo Expectation
Maximization (PEM) algorithm. They have no guarantees of convergence or
convergence to the correct parameters. In this paper, we use Automatic
Differentiation (AD) tools to develop a method, called AD-GMCM, that can
maximize the exact GMCM likelihood. In our simulation studies and experiments
with real data, AD-GMCM finds more accurate parameter estimates than PEM and
yields better performance in clustering and Reproducibility Analysis. We
discuss the advantages of an AD-based approach, to address problems related to
monotonic increase of likelihood and parameter identifiability in GMCM. We also
analyze, for GMCM, two well-known cases of degeneracy of maximum likelihood in
GMM that can lead to spurious clustering solutions. Our analysis shows that,
unlike GMM, GMCM is not affected in one of the cases.
| stat |
Estimating locomotor demands during team play from broadcast-derived
tracking data | The introduction of optical tracking data across sports has given rise to the
ability to dissect athletic performance at a level unfathomable a decade ago.
One specific area that has seen substantial benefit is sports science, as high
resolution coordinate data permits sports scientists to have to-the-second
estimates of external load metrics, such as acceleration load and high speed
running distance, traditionally used to understand the physical toll a game
takes on an athlete. Unfortunately, collecting this data requires installation
of expensive hardware and paying costly licensing fees to data providers,
restricting its availability. Algorithms have been developed that allow a
traditional broadcast feed to be converted to x-y coordinate data, making
tracking data easier to acquire, but coordinates are available for an athlete
only when that player is within the camera frame. Obviously, this leads to
inaccuracies in player load estimates, limiting the usefulness of this data for
sports scientists. In this research, we develop models that predict offscreen
load metrics and demonstrate the viability of broadcast-derived tracking data
for understanding external load in soccer.
| stat |
Commuting Variability by Wage Groups in Baton Rouge 1990-2010 | Residential segregation recently has shifted to more class or income-based in
the United States, and neighborhoods are undergoing significant changes such as
commuting patterns over time. To better understand the commuting inequality
across neighborhoods of different income levels, this research analyzes
commuting variability (in both distance and time) across wage groups as well as
stability over time using the CTPP data 1990-2010 in Baton Rouge. In comparison
to previous work, commuting distance is estimated more accurately by Monte
Carlo simulation of individual trips to mitigate aggregation error and scale
effect. The results based on neighborhoods mean wage rate indicate that
commuting behaviors vary across areas of different wage rates and such
variability is captured by a convex shape. Affluent neighborhoods tended to
commute more but highest-wage neighborhoods retreated for less commuting. This
trend remains relatively stable over time despite an overall transportation
improvement in general. A complementary analysis based on the distribution of
wage groups is conducted to gain more detailed insights and uncovers the
lasting poor mobility (e.g., fewer location and transport options) of the
lowest-wage workers in 1990-2010.
| stat |
Metropolis-Hastings Generative Adversarial Networks | We introduce the Metropolis-Hastings generative adversarial network (MH-GAN),
which combines aspects of Markov chain Monte Carlo and GANs. The MH-GAN draws
samples from the distribution implicitly defined by a GAN's
discriminator-generator pair, as opposed to standard GANs which draw samples
from the distribution defined only by the generator. It uses the discriminator
from GAN training to build a wrapper around the generator for improved
sampling. With a perfect discriminator, this wrapped generator samples from the
true distribution on the data exactly even when the generator is imperfect. We
demonstrate the benefits of the improved generator on multiple benchmark
datasets, including CIFAR-10 and CelebA, using the DCGAN, WGAN, and progressive
GAN.
| stat |
Patterns of Effects and Sensitivity Analysis for
Differences-in-Differences | Applied analysts often use the differences-in-differences (DID) method to
estimate the causal effect of policy interventions with observational data. The
method is widely used, as the required before and after comparison of a treated
and control group is commonly encountered in practice. DID removes bias from
unobserved time-invariant confounders. While DID removes bias from
time-invariant confounders, bias from time-varying confounders may be present.
Hence, like any observational comparison, DID studies remain susceptible to
bias from hidden confounders. Here, we develop a method of sensitivity analysis
that allows investigators to quantify the amount of bias necessary to change a
study's conclusions. Our method operates within a matched design that removes
bias from observed baseline covariates. We develop methods for both binary and
continuous outcomes. We then apply our methods to two different empirical
examples from the social sciences. In the first application, we study the
effect of changes to disability payments in Germany. In the second, we
re-examine whether election day registration increased turnout in Wisconsin.
| stat |
Learning spectrograms with convolutional spectral kernels | We introduce the convolutional spectral kernel (CSK), a novel family of
non-stationary, nonparametric covariance kernels for Gaussian process (GP)
models, derived from the convolution between two imaginary radial basis
functions. We present a principled framework to interpret CSK, as well as other
deep probabilistic models, using approximated Fourier transform, yielding a
concise representation of input-frequency spectrogram. Observing through the
lens of the spectrogram, we provide insight on the interpretability of deep
models. We then infer the functional hyperparameters using scalable variational
and MCMC methods. On small- and medium-sized spatiotemporal datasets, we
demonstrate improved generalization of GP models when equipped with CSK, and
their capability to extract non-stationary periodic patterns.
| stat |
Gaussian Process Manifold Interpolation for Probabilistic Atrial
Activation Maps and Uncertain Conduction Velocity | In patients with atrial fibrillation, local activation time (LAT) maps are
routinely used for characterising patient pathophysiology. The gradient of LAT
maps can be used to calculate conduction velocity (CV), which directly relates
to material conductivity and may provide an important measure of atrial
substrate properties. Including uncertainty in CV calculations would help with
interpreting the reliability of these measurements. Here, we build upon a
recent insight into reduced-rank Gaussian processes (GP) to perform
probabilistic interpolation of uncertain LAT directly on human atrial
manifolds. Our Gaussian Process Manifold Interpolation (GPMI) method accounts
for the topology of the atria, and allows for calculation of statistics for
predicted CV. We demonstrate our method on two clinical cases, and perform
validation against a simulated ground truth. CV uncertainty depends on data
density, wave propagation direction, and CV magnitude. GPMI is suitable for
probabilistic interpolation of other uncertain quantities on non-Euclidean
manifolds.
| stat |
A guide to Value of Information methods for prioritising research in
health impact modelling | Health impact simulation models are used to predict how a proposed
intervention or scenario will affect public health outcomes, based on available
data and knowledge of the process. The outputs of these models are uncertain
due to uncertainty in the structure and inputs to the model. In order to assess
the extent of uncertainty in the outcome we must quantify all potentially
relevant uncertainties. Then to reduce uncertainty we should obtain and analyse
new data, but it may be unclear which parts of the model would benefit from
such extra research.
This paper presents methods for uncertainty quantification and research
prioritisation in health impact models based on Value of Information (VoI)
analysis. Specifically, we
1. discuss statistical methods for quantifying uncertainty in this type of
model, given the typical kinds of data that are available, which are often
weaker than the ideal data that are desired;
2. show how the expected value of partial perfect information (EVPPI) can be
calculated to compare how uncertainty in each model parameter influences
uncertainty in the output;
3. show how research time can be prioritised efficiently, in the light of
which components contribute most to outcome uncertainty.
The same methods can be used whether the purpose of the model is to estimate
quantities of interest to a policy maker, or to explicitly decide between
policies. We demonstrate how these methods might be used in a model of the
impact of air pollution on health outcomes.
| stat |
Structured Sparsity Modeling for Improved Multivariate Statistical
Analysis based Fault Isolation | In order to improve the fault diagnosis capability of multivariate
statistical methods, this article introduces a fault isolation framework based
on structured sparsity modeling. The developed method relies on the
reconstruction based contribution analysis and the process structure
information can be incorporated into the reconstruction objective function in
the form of structured sparsity regularization terms. The structured sparsity
terms allow selection of fault variables over structures like blocks or
networks of process variables, hence more accurate fault isolation can be
achieved. Four structured sparsity terms corresponding to different kinds of
process information are considered, namely, partially known sparse support,
block sparsity, clustered sparsity and tree-structured sparsity. The
optimization problems involving the structured sparsity terms can be solved
using the Alternating Direction Method of Multipliers (ADMM) algorithm, which
is fast and efficient. Through a simulation example and an application study to
a coal-fired power plant, it is verified that the proposed method can better
isolate faulty variables by incorporating process structure information.
| stat |
Deep learning for prediction of complex geology ahead of drilling | During a geosteering operation the well path is intentionally adjusted in
response to the new data acquired while drilling. To achieve consistent
high-quality decisions, especially when drilling in complex environments,
decision support systems can help cope with high volumes of data and
interpretation complexities. They can assimilate the real-time measurements
into a probabilistic earth model and use the updated model for decision
recommendations.
Recently, machine learning (ML) techniques have enabled a wide range of
methods that redistribute computational cost from on-line to off-line
calculations. In this paper, we introduce two ML techniques into the
geosteering decision support framework. Firstly, a complex earth model
representation is generated using a Generative Adversarial Network (GAN).
Secondly, a commercial extra-deep electromagnetic simulator is represented
using a Forward Deep Neural Network (FDNN).
The numerical experiments demonstrate that the combination of the GAN and the
FDNN in an ensemble randomized maximum likelihood data assimilation scheme
provides real-time estimates of complex geological uncertainty. This yields
reduction of geological uncertainty ahead of the drill-bit from the
measurements gathered behind and around the well bore.
| stat |
Clustering small datasets in high-dimension by random projection | Datasets in high-dimension do not typically form clusters in their original
space; the issue is worse when the number of points in the dataset is small. We
propose a low-computation method to find statistically significant clustering
structures in a small dataset. The method proceeds by projecting the data on a
random line and seeking binary clusterings in the resulting one-dimensional
data. Non-linear separations are obtained by extending the feature space using
monomials of higher degrees in the original features. The statistical validity
of the clustering structures obtained is tested in the projected
one-dimensional space, thus bypassing the challenge of statistical validation
in high-dimension. Projecting on a random line is an extreme dimension
reduction technique that has previously been used successfully as part of a
hierarchical clustering method for high-dimensional data. Our experiments show
that with this simplified framework, statistically significant clustering
structures can be found with as few as 100-200 points, depending on the
dataset. The different structures uncovered are found to persist as more points
are added to the dataset.
| stat |
Testing the fit of relational models | Relational models generalize log-linear models to arbitrary discrete sample
spaces by specifying effects associated with any subsets of their cells. A
relational model may include an overall effect, pertaining to every cell after
a reparameterization, and in this case, the properties of the maximum
likelihood estimates (MLEs) are analogous to those computed under traditional
log-linear models, and the goodness-of-fit tests are also the same. If an
overall effect is not present in any reparameterization, the properties of the
MLEs are considerably different, and the Poisson and multinomial MLEs are not
equivalent. In the Poisson case, if the overall effect is not present, the
observed total is not always preserved by the MLE, and thus, the likelihood
ratio statistic is not identical with twice the Kullback-Leibler divergence.
However, as demonstrated, its general form may be obtained from the Bregman
divergence. The asymptotic equivalence of the Pearson chi-squared and
likelihood ratio statistics holds, but the generality considered here requires
extended proofs.
| stat |
Bayesian Group Selection in Logistic Regression with Application to MRI
Data Analysis | We consider Bayesian logistic regression models with group-structured
covariates. In high-dimensional settings, it is often assumed that only small
portion of groups are significant, thus consistent group selection is of
significant importance. While consistent frequentist group selection methods
have been proposed, theoretical properties of Bayesian group selection methods
for logistic regression models have not been investigated yet. In this paper,
we consider a hierarchical group spike and slab prior for logistic regression
models in high-dimensional settings. Under mild conditions, we establish strong
group selection consistency of the induced posterior, which is the first
theoretical result in the Bayesian literature. Through simulation studies, we
demonstrate that the performance of the proposed method outperforms existing
state-of-the-art methods in various settings. We further apply our method to an
MRI data set for predicting Parkinson's disease and show its benefits over
other contenders.
| stat |
Scalable Deep Generative Relational Models with High-Order Node
Dependence | We propose a probabilistic framework for modelling and exploring the latent
structure of relational data. Given feature information for the nodes in a
network, the scalable deep generative relational model (SDREM) builds a deep
network architecture that can approximate potential nonlinear mappings between
nodes' feature information and the nodes' latent representations. Our
contribution is two-fold: (1) We incorporate high-order neighbourhood structure
information to generate the latent representations at each node, which vary
smoothly over the network. (2) Due to the Dirichlet random variable structure
of the latent representations, we introduce a novel data augmentation trick
which permits efficient Gibbs sampling. The SDREM can be used for large sparse
networks as its computational cost scales with the number of positive links. We
demonstrate its competitive performance through improved link prediction
performance on a range of real-world datasets.
| stat |
A Novel Adaptive Kernel for the RBF Neural Networks | In this paper, we propose a novel adaptive kernel for the radial basis
function (RBF) neural networks. The proposed kernel adaptively fuses the
Euclidean and cosine distance measures to exploit the reciprocating properties
of the two. The proposed framework dynamically adapts the weights of the
participating kernels using the gradient descent method thereby alleviating the
need for predetermined weights. The proposed method is shown to outperform the
manual fusion of the kernels on three major problems of estimation namely
nonlinear system identification, pattern classification and function
approximation.
| stat |
Accurate, reliable and fast robustness evaluation | Throughout the past five years, the susceptibility of neural networks to
minimal adversarial perturbations has moved from a peculiar phenomenon to a
core issue in Deep Learning. Despite much attention, however, progress towards
more robust models is significantly impaired by the difficulty of evaluating
the robustness of neural network models. Today's methods are either fast but
brittle (gradient-based attacks), or they are fairly reliable but slow (score-
and decision-based attacks). We here develop a new set of gradient-based
adversarial attacks which (a) are more reliable in the face of gradient-masking
than other gradient-based attacks, (b) perform better and are more query
efficient than current state-of-the-art gradient-based attacks, (c) can be
flexibly adapted to a wide range of adversarial criteria and (d) require
virtually no hyperparameter tuning. These findings are carefully validated
across a diverse set of six different models and hold for L0, L1, L2 and Linf
in both targeted as well as untargeted scenarios. Implementations will soon be
available in all major toolboxes (Foolbox, CleverHans and ART). We hope that
this class of attacks will make robustness evaluations easier and more
reliable, thus contributing to more signal in the search for more robust
machine learning models.
| stat |
The Role of Distributional Overlap on the Precision Gain of Bounds for
Generalization | Over the past ten years, propensity score methods have made an important
contribution to improving generalizations from studies that do not select
samples randomly from a population of inference. However, these methods require
assumptions and recent work has considered the role of bounding approaches that
provide a range of treatment impact estimates that are consistent with the
observable data. An important limitation to bound estimates is that they can be
uninformatively wide. This has motivated research on the use of propensity
score stratification to narrow bounds. This article assesses the role of
distributional overlap in propensity scores on the effectiveness of
stratification to tighten bounds. Using the results of two simulation studies
and two case studies, I evaluate the relationship between distributional
overlap and precision gain and discuss the implications when propensity score
stratification is used as a method to improve precision in the bounding
framework.
| stat |
Tensor Basis Gaussian Process Models of Hyperelastic Materials | In this work, we develop Gaussian process regression (GPR) models of
hyperelastic material behavior. First, we consider the direct approach of
modeling the components of the Cauchy stress tensor as a function of the
components of the Finger stretch tensor in a Gaussian process. We then consider
an improvement on this approach that embeds rotational invariance of the
stress-stretch constitutive relation in the GPR representation. This approach
requires fewer training examples and achieves higher accuracy while maintaining
invariance to rotations exactly. Finally, we consider an approach that recovers
the strain-energy density function and derives the stress tensor from this
potential. Although the error of this model for predicting the stress tensor is
higher, the strain-energy density is recovered with high accuracy from limited
training data. The approaches presented here are examples of physics-informed
machine learning. They go beyond purely data-driven approaches by embedding the
physical system constraints directly into the Gaussian process representation
of materials models.
| stat |
Enhancing single-arm phase II trials by inclusion of matched control
patients | When a novel treatment has successfully passed phase I, different options to
design subsequent phase II trials are available. One approach is a single-arm
trial, comparing the response rate in the intervention group against a fixed
proportion. Another alternative is to conduct a randomized phase II trial,
comparing the new treatment with placebo or the current standard. A significant
problem arises in both approaches when the investigated patient population is
very heterogeneous regarding prognostic factors. For the situation that a
substantial dataset of historical controls exists, we propose an approach to
enhance the classic single-arm trial design by including matched control
patients. The outcome of the observed study population can be adjusted based on
the matched controls with a comparable distribution of known confounders. We
propose an adaptive two-stage design with the options of early stopping for
futility and recalculation of the sample size taking the matching rate, number
of matching partners, and observed treatment effect into account. The
performance of the proposed design in terms of type I error rate, power, and
expected sample size is investigated via simulation studies based on a
hypothetical phase II trial investigating a novel therapy for patients with
acute myeloid leukemia.
| stat |
A Simple Heuristic for Bayesian Optimization with A Low Budget | The aim of black-box optimization is to optimize an objective function within
the constraints of a given evaluation budget. In this problem, it is generally
assumed that the computational cost for evaluating a point is large; thus, it
is important to search efficiently with as low budget as possible. Bayesian
optimization is an efficient method for black-box optimization and provides
exploration-exploitation trade-off by constructing a surrogate model that
considers uncertainty of the objective function. However, because Bayesian
optimization should construct the surrogate model for the entire search space,
it does not exhibit good performance when points are not sampled sufficiently.
In this study, we develop a heuristic method refining the search space for
Bayesian optimization when the available evaluation budget is low. The proposed
method refines a promising region by dividing the original region so that
Bayesian optimization can be executed with the promising region as the initial
search space. We confirm that Bayesian optimization with the proposed method
outperforms Bayesian optimization alone and shows equal or better performance
to two search-space division algorithms through experiments on the benchmark
functions and the hyperparameter optimization of machine learning algorithms.
| stat |
A Unified Taylor Framework for Revisiting Attribution Methods | Attribution methods have been developed to understand the decision-making
process of machine learning models, especially deep neural networks, by
assigning importance scores to individual features. Existing attribution
methods often built upon empirical intuitions and heuristics. There still lacks
a general and theoretical framework that not only can unify these attribution
methods, but also theoretically reveal their rationales, fidelity, and
limitations. To bridge the gap, in this paper, we propose a Taylor attribution
framework and reformulate seven mainstream attribution methods into the
framework. Based on reformulations, we analyze the attribution methods in terms
of rationale, fidelity, and limitation. Moreover, We establish three principles
for a good attribution in the Taylor attribution framework, i.e., low
approximation error, correct contribution assignment, and unbiased baseline
selection. Finally, we empirically validate the Taylor reformulations and
reveal a positive correlation between the attribution performance and the
number of principles followed by the attribution method via benchmarking on
real-world datasets.
| stat |
Bayesian subset selection and variable importance for interpretable
prediction and classification | Subset selection is a valuable tool for interpretable learning, scientific
discovery, and data compression. However, classical subset selection is often
eschewed due to selection instability, computational bottlenecks, and lack of
post-selection inference. We address these challenges from a Bayesian
perspective. Given any Bayesian predictive model $\mathcal{M}$, we elicit
predictively-competitive subsets using linear decision analysis. The approach
is customizable for (local) prediction or classification and provides
interpretable summaries of $\mathcal{M}$. A key quantity is the acceptable
family of subsets, which leverages the predictive distribution from
$\mathcal{M}$ to identify subsets that offer nearly-optimal prediction. The
acceptable family spawns new (co-) variable importance metrics based on whether
variables (co-) appear in all, some, or no acceptable subsets. Crucially, the
linear coefficients for any subset inherit regularization and predictive
uncertainty quantification via $\mathcal{M}$. The proposed approach exhibits
excellent prediction, interval estimation, and variable selection for simulated
data, including $p=400 > n$. These tools are applied to a large education
dataset with highly correlated covariates, where the acceptable family is
especially useful. Our analysis provides unique insights into the combination
of environmental, socioeconomic, and demographic factors that predict
educational outcomes, and features highly competitive prediction with
remarkable stability.
| stat |
On Linear Convergence of Weighted Kernel Herding | We provide a novel convergence analysis of two popular sampling algorithms,
Weighted Kernel Herding and Sequential Bayesian Quadrature, that are used to
approximate the expectation of a function under a distribution. Existing
theoretical analysis was insufficient to explain the empirical successes of
these algorithms. We improve upon existing convergence rates to show that,
under mild assumptions, these algorithms converge linearly. To this end, we
also suggest a simplifying assumption that is true for most cases in finite
dimensions, and that acts as a sufficient condition for linear convergence to
hold in the much harder case of infinite dimensions. When this condition is not
satisfied, we provide a weaker convergence guarantee. Our analysis also yields
a new distributed algorithm for large-scale computation that we prove converges
linearly under the same assumptions. Finally, we provide an empirical
evaluation to test the proposed algorithm for a real world application.
| stat |
Joint Nonparametric Precision Matrix Estimation with Confounding | We consider the problem of precision matrix estimation where, due to
extraneous confounding of the underlying precision matrix, the data are
independent but not identically distributed. While such confounding occurs in
many scientific problems, our approach is inspired by recent neuroscientific
research suggesting that brain function, as measured using functional magnetic
resonance imagine (fMRI), is susceptible to confounding by physiological noise
such as breathing and subject motion. Following the scientific motivation, we
propose a graphical model, which in turn motivates a joint nonparametric
estimator. We provide theoretical guarantees for the consistency and the
convergence rate of the proposed estimator. In addition, we demonstrate that
the optimization of the proposed estimator can be transformed into a series of
linear programming problems, and thus be efficiently solved in parallel.
Empirical results are presented using simulated and real brain imaging data,
which suggest that our approach improves precision matrix estimation, as
compared to baselines, when confounding is present.
| stat |
Stable discovery of interpretable subgroups via calibration in causal
studies | Building on Yu and Kumbier's PCS framework and for randomized experiments, we
introduce a novel methodology for Stable Discovery of Interpretable Subgroups
via Calibration (StaDISC), with large heterogeneous treatment effects. StaDISC
was developed during our re-analysis of the 1999-2000 VIGOR study, an 8076
patient randomized controlled trial (RCT), that compared the risk of adverse
events from a then newly approved drug, Rofecoxib (Vioxx), to that from an
older drug Naproxen. Vioxx was found to, on average and in comparison to
Naproxen, reduce the risk of gastrointestinal (GI) events but increase the risk
of thrombotic cardiovascular (CVT) events. Applying StaDISC, we fit 18 popular
conditional average treatment effect (CATE) estimators for both outcomes and
use calibration to demonstrate their poor global performance. However, they are
locally well-calibrated and stable, enabling the identification of patient
groups with larger than (estimated) average treatment effects. In fact, StaDISC
discovers three clinically interpretable subgroups each for the GI outcome
(totaling 29.4% of the study size) and the CVT outcome (totaling 11.0%).
Complementary analyses of the found subgroups using the 2001-2004 APPROVe
study, a separate independently conducted RCT with 2587 patients, provides
further supporting evidence for the promise of StaDISC.
| stat |
Block Gibbs samplers for logistic mixed models: convergence properties
and a comparison with full Gibbs samplers | Logistic linear mixed model (LLMM) is one of the most widely used statistical
models. Generally, Markov chain Monte Carlo algorithms are used to explore the
posterior densities associated with Bayesian LLMMs. Polson, Scott and Windle's
(2013) Polya-Gamma data augmentation (DA) technique can be used to construct
full Gibbs (FG) samplers for LLMMs. Here, we develop efficient block Gibbs (BG)
samplers for Bayesian LLMMs using the Polya-Gamma DA method. We compare the FG
and BG samplers in the context of simulated and real data examples as the
correlation between the fixed and random effects changes as well as when the
dimensions of the design matrices vary. These numerical examples demonstrate
superior performance of the BG samplers over the FG samplers. We also derive
conditions guaranteeing geometric ergodicity of the BG Markov chain when the
popular improper uniform prior is assigned on the regression coefficients and
proper or improper priors are placed on the variance parameters of the random
effects. This theoretical result has important practical implications as it
justifies the use of asymptotically valid Monte Carlo standard errors for
Markov chain based estimates of posterior quantities.
| stat |
Estimation of Failure Probabilities via Local Subset Approximations | We here consider the subset simulation method which approaches a failure
event using a decreasing sequence of nested intermediate failure events. The
method resembles importance sampling, which actively explores a probability
space by conditioning the next evaluation on the previous evaluations using a
Markov chain Monte Carlo (MCMC) algorithm. A Markov chain typically requires
many steps to estimate the target distribution, which is impractical with
expensive numerical models. Therefore, we propose to approximate each step of a
Markov chain locally with Gaussian process (GP) regression. Benchmark examples
of reliability analysis show that local approximations significantly improve
overall efficiency of subset simulation. They reduce the number of expensive
limit-state evaluations by over $80\%$. However, GP regression becomes
computationally impractical with increasing dimension. Therefore, to make our
use of a GP feasible, we employ the partial least squares (PLS) regression, a
gradient-free reduction method, locally to explore and utilize a
low-dimensional subspace within a Markov chain. Numerical experiments
illustrate a significant computational gain with maintained sufficient
accuracy.
| stat |
Detecting bearish and bullish markets in financial time series using
hierarchical hidden Markov models | Financial markets exhibit alternating periods of rising and falling prices.
Stock traders seeking to make profitable investment decisions have to account
for those trends, where the goal is to accurately predict switches from bullish
towards bearish markets and vice versa. Popular tools for modeling financial
time series are hidden Markov models, where a latent state process is used to
explicitly model switches among different market regimes. In their basic form,
however, hidden Markov models are not capable of capturing both short- and
long-term trends, which can lead to a misinterpretation of short-term price
fluctuations as changes in the long-term trend. In this paper, we demonstrate
how hierarchical hidden Markov models can be used to draw a comprehensive
picture of financial markets, which can contribute to the development of more
sophisticated trading strategies. The feasibility of the suggested approach is
illustrated in two real-data applications, where we model data from two major
stock indices, the Deutscher Aktienindex and the Standard & Poor's 500.
| stat |
SuRF: a New Method for Sparse Variable Selection, with Application in
Microbiome Data Analysis | In this paper, we present a new variable selection method for regression and
classification purposes. Our method, called Subsampling Ranking Forward
selection (SuRF), is based on LASSO penalised regression, subsampling and
forward-selection methods. SuRF offers major advantages over existing variable
selection methods in terms of both sparsity of selected models and model
inference. We provide an R package that can implement our method for
generalized linear models. We apply our method to classification problems from
microbiome data, using a novel agglomeration approach to deal with the special
tree-like correlation structure of the variables. Existing methods arbitrarily
choose a taxonomic level a priori before performing the analysis, whereas by
combining SuRF with these aggregated variables, we are able to identify the key
biomarkers at the appropriate taxonomic level, as suggested by the data. We
present simulations in multiple sparse settings to demonstrate that our
approach performs better than several other popularly used existing approaches
in recovering the true variables. We apply SuRF to two microbiome data sets:
one about prediction of pouchitis and another for identifying samples from two
healthy individuals. We find that SuRF can provide a better or comparable
prediction with other methods while controlling the false positive rate of
variable selection.
| stat |
Provable Robust Classification via Learned Smoothed Densities | Smoothing classifiers and probability density functions with Gaussian kernels
appear unrelated, but in this work, they are unified for the problem of robust
classification. The key building block is approximating the $\textit{energy
function}$ of the random variable $Y=X+N(0,\sigma^2 I_d)$ with a neural network
which we use to formulate the problem of robust classification in terms of
$\widehat{x}(Y)$, the $\textit{Bayes estimator}$ of $X$ given the noisy
measurements $Y$. We introduce $\textit{empirical Bayes smoothed classifiers}$
within the framework of $\textit{randomized smoothing}$ and study it
theoretically for the two-class linear classifier, where we show one can
improve their robustness above $\textit{the margin}$. We test the theory on
MNIST and we show that with a learned smoothed energy function and a linear
classifier we can achieve provable $\ell_2$ robust accuracies that are
competitive with empirical defenses. This setup can be significantly improved
by $\textit{learning}$ empirical Bayes smoothed classifiers with adversarial
training and on MNIST we show that we can achieve provable robust accuracies
higher than the state-of-the-art empirical defenses in a range of radii. We
discuss some fundamental challenges of randomized smoothing based on a
geometric interpretation due to concentration of Gaussians in high dimensions,
and we finish the paper with a proposal for using walk-jump sampling, itself
based on learned smoothed densities, for robust classification.
| stat |
Distance Assisted Recursive Testing | In many applications, a large number of features are collected with the goal
to identify a few important ones. Sometimes, these features lie in a metric
space with a known distance matrix, which partially reflects their
co-importance pattern. Proper use of the distance matrix will boost the power
of identifying important features. Hence, we develop a new multiple testing
framework named the Distance Assisted Recursive Testing (DART). DART has two
stages. In stage 1, we transform the distance matrix into an aggregation tree,
where each node represents a set of features. In stage 2, based on the
aggregation tree, we set up dynamic node hypotheses and perform multiple
testing on the tree. All rejections are mapped back to the features. Under mild
assumptions, the false discovery proportion of DART converges to the desired
level in high probability converging to one. We illustrate by theory and
simulations that DART has superior performance under various models compared to
the existing methods. We applied DART to a clinical trial in the allogeneic
stem cell transplantation study to identify the gut microbiota whose abundance
will be impacted by the after-transplant care.
| stat |
Relative gradient optimization of the Jacobian term in unsupervised deep
learning | Learning expressive probabilistic models correctly describing the data is a
ubiquitous problem in machine learning. A popular approach for solving it is
mapping the observations into a representation space with a simple joint
distribution, which can typically be written as a product of its marginals --
thus drawing a connection with the field of nonlinear independent component
analysis. Deep density models have been widely used for this task, but their
maximum likelihood based training requires estimating the log-determinant of
the Jacobian and is computationally expensive, thus imposing a trade-off
between computation and expressive power. In this work, we propose a new
approach for exact training of such neural networks. Based on relative
gradients, we exploit the matrix structure of neural network parameters to
compute updates efficiently even in high-dimensional spaces; the computational
cost of the training is quadratic in the input size, in contrast with the cubic
scaling of naive approaches. This allows fast training with objective functions
involving the log-determinant of the Jacobian, without imposing constraints on
its structure, in stark contrast to autoregressive normalizing flows.
| stat |
Think Global and Act Local: Bayesian Optimisation over High-Dimensional
Categorical and Mixed Search Spaces | High-dimensional black-box optimisation remains an important yet notoriously
challenging problem. Despite the success of Bayesian optimisation methods on
continuous domains, domains that are categorical, or that mix continuous and
categorical variables, remain challenging. We propose a novel solution -- we
combine local optimisation with a tailored kernel design, effectively handling
high-dimensional categorical and mixed search spaces, whilst retaining sample
efficiency. We further derive convergence guarantee for the proposed approach.
Finally, we demonstrate empirically that our method outperforms the current
baselines on a variety of synthetic and real-world tasks in terms of
performance, computational costs, or both.
| stat |
Fast Exact Bayesian Inference for Sparse Signals in the Normal Sequence
Model | We consider exact algorithms for Bayesian inference with model selection
priors (including spike-and-slab priors) in the sparse normal sequence model.
Because the best existing exact algorithm becomes numerically unstable for
sample sizes over n=500, there has been much attention for alternative
approaches like approximate algorithms (Gibbs sampling, variational Bayes,
etc.), shrinkage priors (e.g. the Horseshoe prior and the Spike-and-Slab LASSO)
or empirical Bayesian methods. However, by introducing algorithmic ideas from
online sequential prediction, we show that exact calculations are feasible for
much larger sample sizes: for general model selection priors we reach n=25000,
and for certain spike-and-slab priors we can easily reach n=100000. We further
prove a de Finetti-like result for finite sample sizes that characterizes
exactly which model selection priors can be expressed as spike-and-slab priors.
The computational speed and numerical accuracy of the proposed methods are
demonstrated in experiments on simulated data, on a differential gene
expression data set, and to compare the effect of multiple hyper-parameter
settings in the beta-binomial prior. In our experimental evaluation we compute
guaranteed bounds on the numerical accuracy of all new algorithms, which shows
that the proposed methods are numerically reliable whereas an alternative based
on long division is not.
| stat |
Learning Dynamics Model in Reinforcement Learning by Incorporating the
Long Term Future | In model-based reinforcement learning, the agent interleaves between model
learning and planning. These two components are inextricably intertwined. If
the model is not able to provide sensible long-term prediction, the executed
planner would exploit model flaws, which can yield catastrophic failures. This
paper focuses on building a model that reasons about the long-term future and
demonstrates how to use this for efficient planning and exploration. To this
end, we build a latent-variable autoregressive model by leveraging recent ideas
in variational inference. We argue that forcing latent variables to carry
future information through an auxiliary task substantially improves long-term
predictions. Moreover, by planning in the latent space, the planner's solution
is ensured to be within regions where the model is valid. An exploration
strategy can be devised by searching for unlikely trajectories under the model.
Our method achieves higher reward faster compared to baselines on a variety of
tasks and environments in both the imitation learning and model-based
reinforcement learning settings.
| stat |
Estimating Treatment Effects with Observed Confounders and Mediators | Given a causal graph, the do-calculus can express treatment effects as
functionals of the observational joint distribution that can be estimated
empirically. Sometimes the do-calculus identifies multiple valid formulae,
prompting us to compare the statistical properties of the corresponding
estimators. For example, the backdoor formula applies when all confounders are
observed and the frontdoor formula applies when an observed mediator transmits
the causal effect. In this paper, we investigate the over-identified scenario
where both confounders and mediators are observed, rendering both estimators
valid. Addressing the linear Gaussian causal model, we derive the finite-sample
variance for both estimators and demonstrate that either estimator can dominate
the other by an unbounded constant factor depending on the model parameters.
Next, we derive an optimal estimator, which leverages all observed variables to
strictly outperform the backdoor and frontdoor estimators. We also present a
procedure for combining two datasets, with confounders observed in one and
mediators in the other. Finally, we evaluate our methods on both simulated data
and the IHDP and JTPA datasets.
| stat |
Learning Models from Data with Measurement Error: Tackling
Underreporting | Measurement error in observational datasets can lead to systematic bias in
inferences based on these datasets. As studies based on observational data are
increasingly used to inform decisions with real-world impact, it is critical
that we develop a robust set of techniques for analyzing and adjusting for
these biases. In this paper we present a method for estimating the distribution
of an outcome given a binary exposure that is subject to underreporting. Our
method is based on a missing data view of the measurement error problem, where
the true exposure is treated as a latent variable that is marginalized out of a
joint model. We prove three different conditions under which the outcome
distribution can still be identified from data containing only error-prone
observations of the exposure. We demonstrate this method on synthetic data and
analyze its sensitivity to near violations of the identifiability conditions.
Finally, we use this method to estimate the effects of maternal smoking and
opioid use during pregnancy on childhood obesity, two import problems from
public health. Using the proposed method, we estimate these effects using only
subject-reported drug use data and substantially refine the range of estimates
generated by a sensitivity analysis-based approach. Further, the estimates
produced by our method are consistent with existing literature on both the
effects of maternal smoking and the rate at which subjects underreport smoking.
| stat |
Asymmetry approach to study for chemotherapy treatment and devices
failure times data using modified Power function distribution with some
modified estimators | In order to improve the already existing models that are used extensively in
bio sciences and applied sciences research, a new class of Weighted Power
function distribution (WPFD) has been proposed with its various properties and
different modifications to be more applicable in real life. We have provided
the mathematical derivations for the new distribution including moments,
incomplete moments, conditional moments, inverse moments, mean residual
function, vitality function, order statistics, mills ratio, information
function, Shannon entropy, Bonferroni and Lorenz curves and quantile function.
We have also characterized the WPFD, based on doubly truncated mean. The aim of
the study is to increase the application of the Power function distribution.
The main feature of the proposed distribution is that there is no induction of
parameters as compare to the other generalization of the distributions, which
are complexed having many parameters. We have used R programming to estimate
the parameters of the new class of WPFD using Maximum Likelihood Method (MLM),
Percentile Estimators (P.E) and their modified estimators. After analyzing the
data, we conclude that the proposed model WPFD performs better in the data sets
while compared to different competitor models.
| stat |
Nonparametric testing of the dependence structure among
points-marks-covariates in spatial point patterns | We investigate testing of the hypothesis of independence between a covariate
and the marks in a marked point process. It would be rather straightforward if
the (unmarked) point process were independent of the covariate and the marks.
In practice, however, such an assumption is questionable, and possible
dependence between the point process and the covariate or the marks may lead to
incorrect conclusions. Hence we propose to investigate the complete dependence
structure in the triangle points-marks-covariates together. We take advantage
of the recent development of the nonparametric random shift methods, namely the
new variance correction approach, and propose tests of the null hypothesis of
independence between the marks and the covariate and between the points and the
covariate. We present a detailed simulation study showing the performance of
the methods, and provide two theorems establishing the appropriate form of the
correction factors for the variance correction. Finally, we illustrate the use
of the proposed methods in two real applications.
| stat |
Improving the Stability of the Knockoff Procedure: Multiple Simultaneous
Knockoffs and Entropy Maximization | The Model-X knockoff procedure has recently emerged as a powerful approach
for feature selection with statistical guarantees. The advantage of knockoff is
that if we have a good model of the features X, then we can identify salient
features without knowing anything about how the outcome Y depends on X. An
important drawback of knockoffs is its instability: running the procedure twice
can result in very different selected features, potentially leading to
different conclusions. Addressing this instability is critical for obtaining
reproducible and robust results. Here we present a generalization of the
knockoff procedure that we call simultaneous multi-knockoffs. We show that
multi-knockoff guarantees false discovery rate (FDR) control, and is
substantially more stable and powerful compared to the standard (single)
knockoff. Moreover we propose a new algorithm based on entropy maximization for
generating Gaussian multi-knockoffs. We validate the improved stability and
power of multi-knockoffs in systematic experiments. We also illustrate how
multi-knockoffs can improve the accuracy of detecting genetic mutations that
are causally linked to phenotypes.
| stat |
Improving Adversarial Robustness of Ensembles with Diversity Training | Deep Neural Networks are vulnerable to adversarial attacks even in settings
where the attacker has no direct access to the model being attacked. Such
attacks usually rely on the principle of transferability, whereby an attack
crafted on a surrogate model tends to transfer to the target model. We show
that an ensemble of models with misaligned loss gradients can provide an
effective defense against transfer-based attacks. Our key insight is that an
adversarial example is less likely to fool multiple models in the ensemble if
their loss functions do not increase in a correlated fashion. To this end, we
propose Diversity Training, a novel method to train an ensemble of models with
uncorrelated loss functions. We show that our method significantly improves the
adversarial robustness of ensembles and can also be combined with existing
methods to create a stronger defense.
| stat |
Parametric Copula-GP model for analyzing multidimensional neuronal and
behavioral relationships | One of the main challenges in current systems neuroscience is the analysis of
high-dimensional neuronal and behavioral data that are characterized by
different statistics and timescales of the recorded variables. We propose a
parametric copula model which separates the statistics of the individual
variables from their dependence structure, and escapes the curse of
dimensionality by using vine copula constructions. We use a Bayesian framework
with Gaussian Process (GP) priors over copula parameters, conditioned on a
continuous task-related variable. We validate the model on synthetic data and
compare its performance in estimating mutual information against the commonly
used non-parametric algorithms.
Our model provides accurate information estimates when the dependencies in
the data match the parametric copulas used in our framework. When the exact
density estimation with a parametric model is not possible, our Copula-GP model
is still able to provide reasonable information estimates, close to the ground
truth and comparable to those obtained with a neural network estimator.
Finally, we apply our framework to real neuronal and behavioral recordings
obtained in awake mice. We demonstrate the ability of our framework to
1) produce accurate and interpretable bivariate models for the analysis of
inter-neuronal noise correlations or behavioral modulations;
2) expand to more than 100 dimensions and measure information content in the
whole-population statistics. These results demonstrate that the Copula-GP
framework is particularly useful for the analysis of complex multidimensional
relationships between neuronal, sensory and behavioral data.
| stat |
A causal exposure response function with local adjustment for
confounding: Estimating health effects of exposure to low levels of ambient
fine particulate matter | The Clean Air Act mandates that the National Ambient Air Quality Standards
(NAAQS) must be routinely assessed to protect populations based on the latest
science. Therefore, researchers should continue to address whether exposure to
levels of air pollution below the NAAQS is harmful to human health. The
contentious nature surrounding environmental regulations urges us to cast this
question within a causal inference framework. Parametric and semi-parametric
regression approaches have been used to estimate the exposure-response (ER)
curve between ambient air pollution and health outcomes. Most of these
approaches are not formulated within a causal framework, adjust for the same
covariates across all levels of exposure, and do not account for model
uncertainty. We introduce a Bayesian framework for the estimation of a causal
ER curve called LERCA (Local Exposure Response Confounding Adjustment), which
allows for different confounders and different strength of confounding at the
different exposure levels; and propagates uncertainty regarding confounders'
selection and the shape of the ER. LERCA provides a principled way of assessing
the covariates' confounding importance at different exposure levels, providing
researchers with information regarding the variables to adjust for in
regression models. Using simulations, we show that state of the art approaches
perform poorly in estimating the ER curve in the presence of local confounding.
LERCA is used to evaluate the relationship between exposure to ambient PM2.5
and cardiovascular hospitalizations for 5,362 zip codes in the US, while
adjusting for a potentially varying set of confounders across the exposure
range. Ambient PM2.5 leads to an increase in cardiovascular hospitalization
rates when focusing at the low exposure range. Our results indicate that there
is no threshold for the effect of PM2.5 on cardiovascular hospitalizations.
| stat |
Approximating posteriors with high-dimensional nuisance parameters via
integrated rotated Gaussian approximation | Posterior computation for high-dimensional data with many parameters can be
challenging. This article focuses on a new method for approximating posterior
distributions of a low- to moderate-dimensional parameter in the presence of a
high-dimensional or otherwise computationally challenging nuisance parameter.
The focus is on regression models and the key idea is to separate the
likelihood into two components through a rotation. One component involves only
the nuisance parameters, which can then be integrated out using a novel type of
Gaussian approximation. We provide theory on approximation accuracy that holds
for a broad class of forms of the nuisance component and priors. Applying our
method to simulated and real data sets shows that it can outperform
state-of-the-art posterior approximation approaches.
| stat |
A statistical Testing Procedure for Validating Class Labels | Motivated by an open problem of validating protein identities in label-free
shotgun proteomics work-flows, we present a testing procedure to validate
class/protein labels using available measurements across instances/peptides.
More generally, we present a solution to the problem of identifying instances
that are deemed, based on some distance (or quasi-distance) measure, as
outliers relative to the subset of instances assigned to the same class. The
proposed procedure is non-parametric and requires no specific distributional
assumption on the measured distances. The only assumption underlying the
testing procedure is that measured distances between instances within the same
class are stochastically smaller than measured distances between instances from
different classes. The test is shown to simultaneously control the Type I and
Type II error probabilities whilst also controlling the overall error
probability of the repeated testing invoked in the validation procedure of
initial class labeling. The theoretical results are supplemented with results
from an extensive numerical study, simulating a typical setup for labeling
validation in proteomics work-flow applications. These results illustrate the
applicability and viability of our method. Even with up to 25% of instances
mislabeled, our testing procedure maintains a high specificity and greatly
reduces the proportion of mislabeled instances.
| stat |
Intrinsic Sliced Wasserstein Distances for Comparing Collections of
Probability Distributions on Manifolds and Graphs | Collections of probability distributions arise in a variety of statistical
applications ranging from user activity pattern analysis to brain connectomics.
In practice these distributions are represented by histograms over diverse
domain types including finite intervals, circles, cylinders, spheres, other
manifolds, and graphs. This paper introduces an approach for detecting
differences between two collections of histograms over such general domains. We
propose the intrinsic slicing construction that yields a novel class of
Wasserstein distances on manifolds and graphs. These distances are Hilbert
embeddable, allowing us to reduce the histogram collection comparison problem
to a more familiar mean testing problem in a Hilbert space. We provide two
testing procedures one based on resampling and another on combining p-values
from coordinate-wise tests. Our experiments in a variety of data settings show
that the resulting tests are powerful and the p-values are well-calibrated.
Example applications to user activity patterns, spatial data, and brain
connectomics are provided.
| stat |
Horvitz-Thompson-like estimation with distance-based detection
probabilities for circular plot sampling of forests | In circular plot sampling, trees within a given distance from the sample plot
location constitute a sample, which is used to infer characteristics of
interest for the forest area. If the sample is collected using a technical
device located at the sampling point, e.g. a terrestrial laser scanner, all
trees of the sample plot cannot be observed because they hide behind each
other. We propose a Horvitz-Thompson-like estimator with distance-based
detection probabilities derived from stochastic geometry for estimation of
population totals such as stem density and basal area in such situation. We
show that our estimator is unbiased for Poisson forests and give estimates of
variance and approximate confidence intervals for the estimator, unlike any
previous methods. We compare the estimator to two previously published
benchmark methods. The comparison is done through a simulation study where
several plots are simulated either from field measured data or different marked
point processes. The simulations show that the estimator produces lower or
comparable error values than the other methods. In the sample plots based on
the field measured data the bias is relatively small - relative mean of errors
for stem density, for example, varying from 0.3 to 2.2 per cent, depending on
the detection condition - and the empirical coverage probabilities of the
approximate confidence intervals are either similar to the nominal levels or
conservative.
| stat |
Improving ERGM Starting Values Using Simulated Annealing | Much of the theory of estimation for exponential family models, which include
exponential-family random graph models (ERGMs) as a special case, is
well-established and maximum likelihood estimates in particular enjoy many
desirable properties. However, in the case of many ERGMs, direct calculation of
MLEs is impossible and therefore methods for approximating MLEs and/or
alternative estimation methods must be employed. Many MLE approximation methods
require alternative estimates as starting points. We discuss one class of such
alternatives here. The MLE satisfies the so-called "likelihood principle,"
unlike the MPLE. This means that different networks may have different MPLEs
even if they have the same sufficient statistics. We exploit this fact here to
search for improved starting values for approximation-based MLE methods. The
method we propose has shown its merit in producing an MLE for a network dataset
and model that had defied estimation using all other known methods.
| stat |
Primal Dual Interpretation of the Proximal Stochastic Gradient Langevin
Algorithm | We consider the task of sampling with respect to a log concave probability
distribution. The potential of the target distribution is assumed to be
composite, \textit{i.e.}, written as the sum of a smooth convex term, and a
nonsmooth convex term possibly taking infinite values. The target distribution
can be seen as a minimizer of the Kullback-Leibler divergence defined on the
Wasserstein space (\textit{i.e.}, the space of probability measures). In the
first part of this paper, we establish a strong duality result for this
minimization problem. In the second part of this paper, we use the duality gap
arising from the first part to study the complexity of the Proximal Stochastic
Gradient Langevin Algorithm (PSGLA), which can be seen as a generalization of
the Projected Langevin Algorithm. Our approach relies on viewing PSGLA as a
primal dual algorithm and covers many cases where the target distribution is
not fully supported. In particular, we show that if the potential is strongly
convex, the complexity of PSGLA is $O(1/\varepsilon^2)$ in terms of the
2-Wasserstein distance. In contrast, the complexity of the Projected Langevin
Algorithm is $O(1/\varepsilon^{12})$ in terms of total variation when the
potential is convex.
| stat |
Approximate Bayesian Computation for Finite Mixture Models | Finite mixture models are used in statistics and other disciplines, but
inference for mixture models is challenging due, in part, to the multimodality
of the likelihood function and the so-called label switching problem. We
propose extensions of the Approximate Bayesian Computation-Population Monte
Carlo (ABC-PMC) algorithm as an alternative framework for inference on finite
mixture models. There are several decisions to make when implementing an
ABC-PMC algorithm for finite mixture models, including the selection of the
kernels used for moving the particles through the iterations, how to address
the label switching problem, and the choice of informative summary statistics.
Examples are presented to demonstrate the performance of the proposed ABC-PMC
algorithm for mixture modeling. The performance of the proposed method is
evaluated in a simulation study and for the popular recessional velocity galaxy
data.
| stat |
Exact slice sampler for Hierarchical Dirichlet Processes | We propose an exact slice sampler for Hierarchical Dirichlet process (HDP)
and its associated mixture models (Teh et al., 2006). Although there are
existing MCMC algorithms for sampling from the HDP, a slice sampler has been
missing from the literature. Slice sampling is well-known for its desirable
properties including its fast mixing and its natural potential for
parallelization. On the other hand, the hierarchical nature of HDPs poses
challenges to adopting a full-fledged slice sampler that automatically
truncates all the infinite measures involved without ad-hoc modifications. In
this work, we adopt the powerful idea of Bayesian variable augmentation to
address this challenge. By introducing new latent variables, we obtain a full
factorization of the joint distribution that is suitable for slice sampling.
Our algorithm has several appealing features such as (1) fast mixing; (2)
remaining exact while allowing natural truncation of the underlying
infinite-dimensional measures, as in (Kalli et al., 2011), resulting in updates
of only a finite number of necessary atoms and weights in each iteration; and
(3) being naturally suited to parallel implementations. The underlying
principle for joint factorization of the full likelihood is simple and can be
applied to many other settings, such as designing sampling algorithms for
general dependent Dirichlet process (DDP) models.
| stat |
Censored Regression for Modelling International Small Arms Trading and
its "Forensic" Use for Exploring Unreported Trades | In this paper we use a censored regression model to investigate data on the
international trade of small arms and ammunition (SAA) provided by the
Norwegian Initiative on Small Arms Transfers (NISAT). Taking a network based
view on the transfers, we not only rely on exogenous covariates but also
estimate endogenous network effects. We apply a spatial autocorrelation (SAR)
model with multiple weight matrices. The likelihood is maximized employing the
Monte Carlo Expectation Maximization (MCEM) algorithm. Our approach reveals
strong and stable endogenous network effects. Furthermore, we find evidence for
a substantial path dependence as well as a close connection between exports of
civilian and military small arms. The model is then used in a "forensic" manner
to analyse latent network structures and thereby to identify countries with
higher or lower tendency to export or import than reflected in the data. The
approach is also validated using a simulation study.
| stat |
Speech Recognition: Keyword Spotting Through Image Recognition | The problem of identifying voice commands has always been a challenge due to
the presence of noise and variability in speed, pitch, etc. We will compare the
efficacies of several neural network architectures for the speech recognition
problem. In particular, we will build a model to determine whether a one second
audio clip contains a particular word (out of a set of 10), an unknown word, or
silence. The models to be implemented are a CNN recommended by the Tensorflow
Speech Recognition tutorial, a low-latency CNN, and an adversarially trained
CNN. The result is a demonstration of how to convert a problem in audio
recognition to the better-studied domain of image classification, where the
powerful techniques of convolutional neural networks are fully developed.
Additionally, we demonstrate the applicability of the technique of Virtual
Adversarial Training (VAT) to this problem domain, functioning as a powerful
regularizer with promising potential future applications.
| stat |
Device Heterogeneity in Federated Learning: A Superquantile Approach | We propose a federated learning framework to handle heterogeneous client
devices which do not conform to the population data distribution. The approach
hinges upon a parameterized superquantile-based objective, where the parameter
ranges over levels of conformity. We present an optimization algorithm and
establish its convergence to a stationary point. We show how to practically
implement it using secure aggregation by interleaving iterations of the usual
federated averaging method with device filtering. We conclude with numerical
experiments on neural networks as well as linear models on tasks from computer
vision and natural language processing.
| stat |
Stage I non-small cell lung cancer stratification by using a model-based
clustering algorithm with covariates | Lung cancer is currently the leading cause of cancer deaths. Among various
subtypes, the number of patients diagnosed with stage I non-small cell lung
cancer (NSCLC), particularly adenocarcinoma, has been increasing. It is
estimated that 30 - 40\% of stage I patients will relapse, and 10 - 30\% will
die due to recurrence, clearly suggesting the presence of a subgroup that could
be benefited by additional therapy. We hypothesize that current attempts to
identify stage I NSCLC subgroup failed due to covariate effects, such as the
age at diagnosis and differentiation, which may be masking the results. In this
context, to stratify stage I NSCLC, we propose CEM-Co, a model-based clustering
algorithm that removes/minimizes the effects of undesirable covariates during
the clustering process. We applied CEM-Co on a gene expression data set
composed of 129 subjects diagnosed with stage I NSCLC and successfully
identified a subgroup with a significantly different phenotype (poor
prognosis), while standard clustering algorithms failed.
| stat |
On Mixup Training: Improved Calibration and Predictive Uncertainty for
Deep Neural Networks | Mixup~\cite{zhang2017mixup} is a recently proposed method for training deep
neural networks where additional samples are generated during training by
convexly combining random pairs of images and their associated labels. While
simple to implement, it has been shown to be a surprisingly effective method of
data augmentation for image classification: DNNs trained with mixup show
noticeable gains in classification performance on a number of image
classification benchmarks. In this work, we discuss a hitherto untouched aspect
of mixup training -- the calibration and predictive uncertainty of models
trained with mixup. We find that DNNs trained with mixup are significantly
better calibrated -- i.e., the predicted softmax scores are much better
indicators of the actual likelihood of a correct prediction -- than DNNs
trained in the regular fashion. We conduct experiments on a number of image
classification architectures and datasets -- including large-scale datasets
like ImageNet -- and find this to be the case. Additionally, we find that
merely mixing features does not result in the same calibration benefit and that
the label smoothing in mixup training plays a significant role in improving
calibration. Finally, we also observe that mixup-trained DNNs are less prone to
over-confident predictions on out-of-distribution and random-noise data. We
conclude that the typical overconfidence seen in neural networks, even on
in-distribution data is likely a consequence of training with hard labels,
suggesting that mixup be employed for classification tasks where predictive
uncertainty is a significant concern.
| stat |
A general method for regularizing tensor decomposition methods via
pseudo-data | Tensor decomposition methods allow us to learn the parameters of latent
variable models through decomposition of low-order moments of data. A
significant limitation of these algorithms is that there exists no general
method to regularize them, and in the past regularization has mostly been
performed using bespoke modifications to the algorithms, tailored for the
particular form of the desired regularizer. We present a general method of
regularizing tensor decomposition methods which can be used for any likelihood
model that is learnable using tensor decomposition methods and any
differentiable regularization function by supplementing the training data with
pseudo-data. The pseudo-data is optimized to balance two terms: being as close
as possible to the true data and enforcing the desired regularization. On
synthetic, semi-synthetic and real data, we demonstrate that our method can
improve inference accuracy and regularize for a broad range of goals including
transfer learning, sparsity, interpretability, and orthogonality of the learned
parameters.
| stat |
Robust Adaptive Control Charts | In statistical process control, procedures are applied that require
relatively strict conditions for their use. If such assumptions are violated,
these methods become inefficient, leading to increased incidence of false
signals. Therefore, a robust version of control charts is sought to be less
sensitive with respect to a breach of normality and independence in
measurements. Robust control charts, however, usually increase the delay in the
detection of assignable causes. This negative effect can, to some extent, be
removed with the aid of an adaptive approach.
| stat |
Weibull Racing Survival Analysis for Competing Events and a Study of
Loan Payoff and Default | We propose Bayesian nonparametric Weibull delegate racing (WDR) to explicitly
model surviving under competing events and to interpret how the covariates
accelerate or decelerate the event times. WDR explains non-monotonic covariate
effects by racing a potentially infinite number of sub-events, relaxing the
ubiquitous proportional-hazards assumption which may be too restrictive. WDR
can handle different types of censoring and missing event times or types. For
inference, we develop a Gibbs-sampler-based MCMC algorithm along with a maximum
a posteriori estimation for big data applications. We use synthetic data
analysis to demonstrate the flexibility and parsimonious nonlinearity of WDR.
We also use a data set of time to loan payoff and default from Prosper.com to
showcase the interpretability.
| stat |
Estimating Robot Strengths with Application to Selection of Alliance
Members in FIRST Robotics Competitions | Since the inception of the FIRST Robotics Competition (FRC) and its special
playoff system, robotics teams have longed to appropriately quantify the
strengths of their designed robots. The FRC includes a playground draft-like
phase (alliance selection), arguably the most game-changing part of the
competition, in which the top-8 robotics teams in a tournament based on the
FRC's ranking system assess potential alliance members for the opportunity of
partnering in a playoff stage. In such a three-versus-three competition,
several measures and models have been used to characterize actual or relative
robot strengths. However, existing models are found to have poor predictive
performance due to their imprecise estimates of robot strengths caused by a
small ratio of the number of observations to the number of robots. A more
general regression model with latent clusters of robot strengths is, thus,
proposed to enhance their predictive capacities. Two effective estimation
procedures are further developed to simultaneously estimate the number of
clusters, clusters of robots, and robot strengths. Meanwhile, some measures are
used to assess the predictive ability of competing models, the agreement
between published FRC measures of strength and model-based robot strengths of
all, playoff, and FRC top-8 robots, and the agreement between FRC top-8 robots
and model-based top robots. Moreover, the stability of estimated robot
strengths and accuracies is investigated to determine whether the scheduled
matches are excessive or insufficient. In the analysis of qualification data
from the 2018 FRC Houston and Detroit championships, the predictive ability of
our model is also shown to be significantly better than those of existing
models. Teams who adopt the new model can now appropriately rank their
preferences for playoff alliance partners with greater predictive capability
than before.
| stat |
Interaction Matters: A Note on Non-asymptotic Local Convergence of
Generative Adversarial Networks | Motivated by the pursuit of a systematic computational and algorithmic
understanding of Generative Adversarial Networks (GANs), we present a simple
yet unified non-asymptotic local convergence theory for smooth two-player
games, which subsumes several discrete-time gradient-based saddle point
dynamics. The analysis reveals the surprising nature of the off-diagonal
interaction term as both a blessing and a curse. On the one hand, this
interaction term explains the origin of the slow-down effect in the convergence
of Simultaneous Gradient Ascent (SGA) to stable Nash equilibria. On the other
hand, for the unstable equilibria, exponential convergence can be proved thanks
to the interaction term, for four modified dynamics proposed to stabilize GAN
training: Optimistic Mirror Descent (OMD), Consensus Optimization (CO),
Implicit Updates (IU) and Predictive Method (PM). The analysis uncovers the
intimate connections among these stabilizing techniques, and provides detailed
characterization on the choice of learning rate. As a by-product, we present a
new analysis for OMD proposed in Daskalakis, Ilyas, Syrgkanis, and Zeng [2017]
with improved rates.
| stat |
Conjugate Nearest Neighbor Gaussian Process Models for Efficient
Statistical Interpolation of Large Spatial Data | A key challenge in spatial statistics is the analysis for massive
spatially-referenced data sets. Such analyses often proceed from Gaussian
process specifications that can produce rich and robust inference, but involve
dense covariance matrices that lack computationally exploitable structures. The
matrix computations required for fitting such models involve floating point
operations in cubic order of the number of spatial locations and dynamic memory
storage in quadratic order. Recent developments in spatial statistics offer a
variety of massively scalable approaches. Bayesian inference and hierarchical
models, in particular, have gained popularity due to their richness and
flexibility in accommodating spatial processes. Our current contribution is to
provide computationally efficient exact algorithms for spatial interpolation of
massive data sets using scalable spatial processes. We combine low-rank
Gaussian processes with efficient sparse approximations. Following recent work
by [1], we model the low-rank process using a Gaussian predictive process (GPP)
and the residual process as a sparsity-inducing nearest-neighbor Gaussian
process (NNGP). A key contribution here is to implement these models using
exact conjugate Bayesian modeling to avoid expensive iterative algorithms.
Through the simulation studies, we evaluate performance of the proposed
approach and the robustness of our models, especially for long range
prediction. We implement our approaches for remotely sensed light detection and
ranging (LiDAR) data collected over the US Forest Service Tanana Inventory Unit
(TIU) in a remote portion of Interior Alaska.
| stat |
Covariate balancing for causal inference on categorical and continuous
treatments | We propose novel estimators for categorical and continuous treatments by
using an optimal covariate balancing strategy for inverse probability
weighting. The resulting estimators are shown to be consistent and
asymptotically normal for causal contrasts of interest, either when the model
explaining treatment assignment is correctly specified, or when the correct set
of bases for the outcome models has been chosen and the assignment model is
sufficiently rich. For the categorical treatment case, we show that the
estimator attains the semiparametric efficiency bound when all models are
correctly specified. For the continuous case, the causal parameter of interest
is a function of the treatment dose. The latter is not parametrized and the
estimators proposed are shown to have bias and variance of the classical
nonparametric rate. Asymptotic results are complemented with simulations
illustrating the finite sample properties. Our analysis of a data set suggests
a nonlinear effect of BMI on the decline in self reported health.
| stat |
Nonclosedness of Sets of Neural Networks in Sobolev Spaces | We examine the closedness of sets of realized neural networks of a fixed
architecture in Sobolev spaces. For an exactly $m$-times differentiable
activation function $\rho$, we construct a sequence of neural networks
$(\Phi_n)_{n \in \mathbb{N}}$ whose realizations converge in order-$(m-1)$
Sobolev norm to a function that cannot be realized exactly by a neural network.
Thus, sets of realized neural networks are not closed in order-$(m-1)$ Sobolev
spaces $W^{m-1,p}$ for $p \in [1,\infty]$. We further show that these sets are
not closed in $W^{m,p}$ under slightly stronger conditions on the $m$-th
derivative of $\rho$. For a real analytic activation function, we show that
sets of realized neural networks are not closed in $W^{k,p}$ for any $k \in
\mathbb{N}$. The nonclosedness allows for approximation of non-network target
functions with unbounded parameter growth. We partially characterize the rate
of parameter growth for most activation functions by showing that a specific
sequence of realized neural networks can approximate the activation function's
derivative with weights increasing inversely proportional to the $L^p$
approximation error. Finally, we present experimental results showing that
networks are capable of closely approximating non-network target functions with
increasing parameters via training.
| stat |
Kernel Interpolation for Scalable Online Gaussian Processes | Gaussian processes (GPs) provide a gold standard for performance in online
settings, such as sample-efficient control and black box optimization, where we
need to update a posterior distribution as we acquire data in a sequential
fashion. However, updating a GP posterior to accommodate even a single new
observation after having observed $n$ points incurs at least $O(n)$
computations in the exact setting. We show how to use structured kernel
interpolation to efficiently recycle computations for constant-time $O(1)$
online updates with respect to the number of points $n$, while retaining exact
inference. We demonstrate the promise of our approach in a range of online
regression and classification settings, Bayesian optimization, and active
sampling to reduce error in malaria incidence forecasting. Code is available at
https://github.com/wjmaddox/online_gp.
| stat |
Deep Mixed Effect Model using Gaussian Processes: A Personalized and
Reliable Prediction for Healthcare | We present a personalized and reliable prediction model for healthcare, which
can provide individually tailored medical services such as diagnosis, disease
treatment, and prevention. Our proposed framework targets at making
personalized and reliable predictions from time-series data, such as Electronic
Health Records (EHR), by modeling two complementary components: i) a shared
component that captures global trend across diverse patients and ii) a
patient-specific component that models idiosyncratic variability for each
patient. To this end, we propose a composite model of a deep neural network to
learn complex global trends from the large number of patients, and Gaussian
Processes (GP) to probabilistically model individual time-series given
relatively small number of visits per patient. We evaluate our model on diverse
and heterogeneous tasks from EHR datasets and show practical advantages over
standard time-series deep models such as pure Recurrent Neural Network (RNN).
| stat |
Higher-order interactions in statistical physics and machine learning: A
model-independent solution to the inverse problem at equilibrium | The problem of inferring pair-wise and higher-order interactions in complex
systems involving large numbers of interacting variables, from observational
data, is fundamental to many fields. Known to the statistical physics community
as the inverse problem, it has become accessible in recent years due to real
and simulated 'big' data being generated. Current approaches to the inverse
problem rely on parametric assumptions, physical approximations, e.g.
mean-field theory, and ignoring higher-order interactions which may lead to
biased or incorrect estimates. We bypass these shortcomings using a
cross-disciplinary approach and demonstrate that none of these assumptions and
approximations are necessary: We introduce a universal, model-independent, and
fundamentally unbiased estimator of all-order symmetric interactions, via the
non-parametric framework of Targeted Learning, a subfield of mathematical
statistics. Due to its universality, our definition is readily applicable to
any system at equilibrium with binary and categorical variables, be it magnetic
spins, nodes in a neural network, or protein networks in biology. Our approach
is targeted, not requiring fitting unnecessary parameters. Instead, it expends
all data on estimating interactions, hence substantially increasing accuracy.
We demonstrate the generality of our technique both analytically and
numerically on (i) the 2-dimensional Ising model, (ii) an Ising-like model with
4-point interactions, (iii) the Restricted Boltzmann Machine, and (iv)
simulated individual-level human DNA variants and representative traits. The
latter demonstrates the applicability of this approach to discover epistatic
interactions causal of disease in population biomedicine.
| stat |
A Note on Debiased/Double Machine Learning Logistic Partially Linear
Model | It is of particular interests in many application fields to draw doubly
robust inference of a logistic partially linear model with the predictor
specified as combination of a targeted low dimensional linear parametric
function and a nuisance nonparametric function. In recent, Tan (2019) proposed
a simple and flexible doubly robust estimator for this purpose. They introduced
the two nuisance models, i.e. nonparametric component in the logistic model and
conditional mean of the exposure covariates given the other covariates and
fixed response, and specified them as fixed dimensional parametric models.
Their framework could be potentially extended to machine learning or high
dimensional nuisance modelling exploited recently, e.g. in Chernozhukovet al.
(2018a,b) and Smucler et al. (2019); Tan (2020). Motivated by this, we derive
the debiased/double machine learning logistic partially linear model in this
note. For construction of the nuisance models, we separately consider the use
of high dimensional sparse parametric models and general machine learning
methods. By deriving certain moment equations to calibrate the first order bias
of the nuisance models, we preserve a model double robustness property on high
dimensional ultra-sparse nuisance models. We also discuss and compare the
underlying assumption of our method with debiased LASSO (Van deGeer et al.,
2014). To implement the machine learning proposal, we design a full model
refitting procedure that allows the use of any blackbox conditional mean
estimation method in our framework. Under the machine learning setting, our
method is rate doubly robust in a similar sense as Chernozhukov et al. (2018a).
| stat |
Phenotyping OSA: a time series analysis using fuzzy clustering and
persistent homology | Sleep apnea is a disorder that has serious consequences for the pediatric
population. There has been recent concern that traditional diagnosis of the
disorder using the apnea-hypopnea index may be ineffective in capturing its
multi-faceted outcomes. In this work, we take a first step in addressing this
issue by phenotyping patients using a clustering analysis of airflow time
series. This is approached in three ways: using feature-based fuzzy clustering
in the time and frequency domains, and using persistent homology to study the
signal from a topological perspective. The fuzzy clusters are analyzed in a
novel manner using a Dirichlet regression analysis, while the topological
approach leverages Takens embedding theorem to study the periodicity properties
of the signals.
| stat |
Efficient surrogate modeling methods for large-scale Earth system models
based on machine learning techniques | Improving predictive understanding of Earth system variability and change
requires data-model integration. Efficient data-model integration for complex
models requires surrogate modeling to reduce model evaluation time. However,
building a surrogate of a large-scale Earth system model (ESM) with many output
variables is computationally intensive because it involves a large number of
expensive ESM simulations. In this effort, we propose an efficient surrogate
method capable of using a few ESM runs to build an accurate and
fast-to-evaluate surrogate system of model outputs over large spatial and
temporal domains. We first use singular value decomposition to reduce the
output dimensions, and then use Bayesian optimization techniques to generate an
accurate neural network surrogate model based on limited ESM simulation
samples. Our machine learning based surrogate methods can build and evaluate a
large surrogate system of many variables quickly. Thus, whenever the quantities
of interest change such as a different objective function, a new site, and a
longer simulation time, we can simply extract the information of interest from
the surrogate system without rebuilding new surrogates, which significantly
saves computational efforts. We apply the proposed method to a regional
ecosystem model to approximate the relationship between 8 model parameters and
42660 carbon flux outputs. Results indicate that using only 20 model
simulations, we can build an accurate surrogate system of the 42660 variables,
where the consistency between the surrogate prediction and actual model
simulation is 0.93 and the mean squared error is 0.02. This highly-accurate and
fast-to-evaluate surrogate system will greatly enhance the computational
efficiency in data-model integration to improve predictions and advance our
understanding of the Earth system.
| stat |
Clustering Multivariate Data using Factor Analytic Bayesian Mixtures
with an Unknown Number of Components | Recent work on overfitting Bayesian mixtures of distributions offers a
powerful framework for clustering multivariate data using a latent Gaussian
model which resembles the factor analysis model. The flexibility provided by
overfitting mixture models yields a simple and efficient way in order to
estimate the unknown number of clusters and model parameters by Markov chain
Monte Carlo (MCMC) sampling. The present study extends this approach by
considering a set of eight parameterizations, giving rise to parsimonious
representations of the covariance matrix per cluster. A Gibbs sampler combined
with a prior parallel tempering scheme is implemented in order to approximately
sample from the posterior distribution of the overfitting mixture. The
parameterization and number of factors is selected according to the Bayesian
Information Criterion. Identifiability issues related to label switching are
dealt by post-processing the simulated output with the Equivalence Classes
Representatives algorithm. The contributed method and software are demonstrated
and compared to similar models estimated using the Expectation-Maximization
algorithm on simulated and real datasets. The software is available online at
https://CRAN.R-project.org/package=fabMix.
| stat |
A Distribution-Free Test of Covariate Shift Using Conformal Prediction | Covariate shift is a common and important assumption in transfer learning and
domain adaptation to treat the distributional difference between the training
and testing data. We propose a nonparametric test of covariate shift using the
conformal prediction framework. The construction of our test statistic combines
recent developments in conformal prediction with a novel choice of conformity
score, resulting in a valid and powerful test statistic under very general
settings. To our knowledge, this is the first successful attempt of using
conformal prediction for testing statistical hypotheses. Our method is suitable
for modern machine learning scenarios where the data has high dimensionality
and large sample sizes, and can be effectively combined with existing
classification algorithms to find good conformity score functions. The
performance of the proposed method is demonstrated in synthetic and real data
examples.
| stat |
Defining Estimands Using a Mix of Strategies to Handle Intercurrent
Events in Clinical Trials | Randomized controlled trials (RCT) are the gold standard for evaluation of
the efficacy and safety of investigational interventions. If every patient in
an RCT were to adhere to the randomized treatment, one could simply analyze the
complete data to infer the treatment effect. However, intercurrent events
(ICEs) including the use of concomitant medication for unsatisfactory efficacy,
treatment discontinuation due to adverse events, or lack of efficacy, may lead
to interventions that deviate from the original treatment assignment.
Therefore, defining the appropriate estimand (the appropriate parameter to be
estimated) based on the primary objective of the study is critical prior to
determining the statistical analysis method and analyzing the data. The
International Council for Harmonisation (ICH) E9 (R1), published on November
20, 2019, provided 5 strategies to define the estimand: treatment policy,
hypothetical, composite variable, while on treatment and principal stratum. In
this article, we propose an estimand using a mix of strategies in handling
ICEs. This estimand is an average of the null treatment difference for those
with ICEs potentially related to safety and the treatment difference for the
other patients if they would complete the assigned treatments. Two examples
from clinical trials evaluating anti-diabetes treatments are provided to
illustrate the estimation of this proposed estimand and to compare it with the
estimates for estimands using hypothetical and treatment policy strategies in
handling ICEs.
| stat |
Max-infinitely divisible models and inference for spatial extremes | For many environmental processes, recent studies have shown that the
dependence strength is decreasing when quantile levels increase. This implies
that the popular max-stable models are inadequate to capture the rate of joint
tail decay, and to estimate joint extremal probabilities beyond observed
levels. We here develop a more flexible modeling framework based on the class
of max-infinitely divisible processes, which extend max-stable processes while
retaining dependence properties that are natural for maxima. We propose two
parametric constructions for max-infinitely divisible models, which relax the
max-stability property but remain close to some popular max-stable models
obtained as special cases. The first model considers maxima over a finite,
random number of independent observations, while the second model generalizes
the spectral representation of max-stable processes. Inference is performed
using a pairwise likelihood. We illustrate the benefits of our new modeling
framework on Dutch wind gust maxima calculated over different time units.
Results strongly suggest that our proposed models outperform other natural
models, such as the Student-t copula process and its max-stable limit, even for
large block sizes.
| stat |
Generalized simultaneous component analysis of binary and quantitative
data | In the current era of systems biological research there is a need for the
integrative analysis of binary and quantitative genomics data sets measured on
the same objects. One standard tool of exploring the underlying dependence
structure present in multiple quantitative data sets is simultaneous component
analysis (SCA) model. However, it does not have any provisions when a part of
the data are binary. To this end, we propose the generalized SCA (GSCA) model,
which takes into account the distinct mathematical properties of binary and
quantitative measurements in the maximum likelihood framework. Like in the SCA
model, a common low dimensional subspace is assumed to represent the shared
information between these two distinct types of measurements. However, the GSCA
model can easily be overfitted when a rank larger than one is used, leading to
some of the estimated parameters to become very large. To achieve a low rank
solution and combat overfitting, we propose to use a concave variant of the
nuclear norm penalty. An efficient majorization algorithm is developed to fit
this model with different concave penalties. Realistic simulations (low
signal-to-noise ratio and highly imbalanced binary data) are used to evaluate
the performance of the proposed model in recovering the underlying structure.
Also, a missing value based cross validation procedure is implemented for model
selection. We illustrate the usefulness of the GSCA model for exploratory data
analysis of quantitative gene expression and binary copy number aberration
(CNA) measurements obtained from the GDSC1000 data sets.
| stat |
Res-embedding for Deep Learning Based Click-Through Rate Prediction
Modeling | Recently, click-through rate (CTR) prediction models have evolved from
shallow methods to deep neural networks. Most deep CTR models follow an
Embedding\&MLP paradigm, that is, first mapping discrete id features, e.g. user
visited items, into low dimensional vectors with an embedding module, then
learn a multi-layer perception (MLP) to fit the target. In this way, embedding
module performs as the representative learning and plays a key role in the
model performance. However, in many real-world applications, deep CTR model
often suffers from poor generalization performance, which is mostly due to the
learning of embedding parameters. In this paper, we model user behavior using
an interest delay model, study carefully the embedding mechanism, and obtain
two important results: (i) We theoretically prove that small aggregation radius
of embedding vectors of items which belongs to a same user interest domain will
result in good generalization performance of deep CTR model. (ii) Following our
theoretical analysis, we design a new embedding structure named res-embedding.
In res-embedding module, embedding vector of each item is the sum of two
components: (i) a central embedding vector calculated from an item-based
interest graph (ii) a residual embedding vector with its scale to be relatively
small. Empirical evaluation on several public datasets demonstrates the
effectiveness of the proposed res-embedding structure, which brings significant
improvement on the model performance.
| stat |
Adaptive Algorithm for Sparse Signal Recovery | Spike and slab priors play a key role in inducing sparsity for sparse signal
recovery. The use of such priors results in hard non-convex and mixed integer
programming problems. Most of the existing algorithms to solve the optimization
problems involve either simplifying assumptions, relaxations or high
computational expenses. We propose a new adaptive alternating direction method
of multipliers (AADMM) algorithm to directly solve the presented optimization
problem. The algorithm is based on the one-to-one mapping property of the
support and non-zero element of the signal. At each step of the algorithm, we
update the support by either adding an index to it or removing an index from it
and use the alternating direction method of multipliers to recover the signal
corresponding to the updated support. Experiments on synthetic data and
real-world images show that the proposed AADMM algorithm provides superior
performance and is computationally cheaper, compared to the recently developed
iterative convex refinement (ICR) algorithm.
| stat |
Subsets and Splits
No saved queries yet
Save your SQL queries to embed, download, and access them later. Queries will appear here once saved.