title
stringlengths 6
244
| abstract
stringlengths 19
5.09k
| label
stringclasses 10
values |
|---|---|---|
Unsupervised parameter selection for denoising with the elastic net | Despite recent advances in regularisation theory, the issue of parameter
selection still remains a challenge for most applications. In a recent work the
framework of statistical learning was used to approximate the optimal Tikhonov
regularisation parameter from noisy data. In this work, we improve their
results and extend the analysis to the elastic net regularisation, providing
explicit error bounds on the accuracy of the approximated parameter and the
corresponding regularisation solution in a simplified case. Furthermore, in the
general case we design a data-driven, automated algorithm for the computation
of an approximate regularisation parameter. Our analysis combines statistical
learning theory with insights from regularisation theory. We compare our
approach with state-of-the-art parameter selection criteria and illustrate its
superiority in terms of accuracy and computational time on simulated and real
data sets.
| stat |
Robust Optimization and Inference on Manifolds | We propose a robust and scalable procedure for general optimization and
inference problems on manifolds leveraging the classical idea of
`median-of-means' estimation. This is motivated by ubiquitous examples and
applications in modern data science in which a statistical learning problem can
be cast as an optimization problem over manifolds. Being able to incorporate
the underlying geometry for inference while addressing the need for robustness
and scalability presents great challenges. We address these challenges by first
proving a key lemma that characterizes some crucial properties of geometric
medians on manifolds. In turn, this allows us to prove robustness and tighter
concentration of our proposed final estimator in a subsequent theorem. This
estimator aggregates a collection of subset estimators by taking their
geometric median over the manifold. We illustrate bounds on this estimator via
calculations in explicit examples. The robustness and scalability of the
procedure is illustrated in numerical examples on both simulated and real data
sets.
| stat |
Simultaneous Inference for Pairwise Graphical Models with Generalized
Score Matching | Probabilistic graphical models provide a flexible yet parsimonious framework
for modeling dependencies among nodes in networks. There is a vast literature
on parameter estimation and consistent model selection for graphical models.
However, in many of the applications, scientists are also interested in
quantifying the uncertainty associated with the estimated parameters and
selected models, which current literature has not addressed thoroughly. In this
paper, we propose a novel estimator for statistical inference on edge
parameters in pairwise graphical models based on generalized Hyv\"arinen
scoring rule. Hyv\"arinen scoring rule is especially useful in cases where the
normalizing constant cannot be obtained efficiently in a closed form, which is
a common problem for graphical models, including Ising models and truncated
Gaussian graphical models. Our estimator allows us to perform statistical
inference for general graphical models whereas the existing works mostly focus
on statistical inference for Gaussian graphical models where finding
normalizing constant is computationally tractable. Under mild conditions that
are typically assumed in the literature for consistent estimation, we prove
that our proposed estimator is $\sqrt{n}$-consistent and asymptotically normal,
which allows us to construct confidence intervals and build hypothesis tests
for edge parameters. Moreover, we show how our proposed method can be applied
to test hypotheses that involve a large number of model parameters
simultaneously. We illustrate validity of our estimator through extensive
simulation studies on a diverse collection of data-generating processes.
| stat |
Partial Least Squares for Functional Joint Models | Many biomedical studies have identified important imaging biomarkers that are
associated with both repeated clinical measures and a survival outcome. The
functional joint model (FJM) framework, proposed in Li and Luo (2017),
investigates the association between repeated clinical measures and survival
data, while adjusting for both high-dimensional images and low-dimensional
covariates based upon the functional principal component analysis (FPCA). In
this paper, we propose a novel algorithm for the estimation of FJM based on the
functional partial least squares (FPLS). Our numerical studies demonstrate
that, compared to FPCA, the proposed FPLS algorithm can yield more accurate and
robust estimation and prediction performance in many important scenarios. We
apply the proposed FPLS algorithm to the Alzheimer's Disease Neuroimaging
Initiative (ADNI) study. Data used in the preparation of this article were
obtained from the ADNI database.
| stat |
Tensor Mixed Effects Model with Applications in Nanomanufacturing
Inspection | Raman mapping technique has been used to perform in-line quality inspections
of nanomanufacturing processes. In such an application, massive
high-dimensional Raman mapping data with mixed effects is generated. In
general, fixed effects and random effects in the multi-array Raman data are
associated with different quality characteristics such as fabrication
consistency, uniformity, defects, et al. The existing tensor decomposition
methods cannot separate mixed effects, and existing mixed effects model can
only handle matrix data but not high-dimensional multi-array data. In this
paper, we propose a tensor mixed effects (TME) model to analyze massive
high-dimensional Raman mapping data with complex structure. The proposed TME
model can (i) separate fixed effects and random effects in a tensor domain;
(ii) explore the correlations along different dimensions; and (iii) realize
efficient parameter estimation by a proposed iterative double Flip-Flop
algorithm. We also investigate the properties of the TME model, existence and
identifiability of parameter estimation. The numerical analysis demonstrates
the efficiency and accuracy of the parameter estimation in the TME model.
Convergence and asymptotic properties are discussed in the simulation and
surrogate data analysis. The case study shows an application of the TME model
in quantifying the influence of alignment on carbon nanotubes buckypaper.
Moreover, the TME model can be applied to provide potential solutions for a
family of tensor data analytics problems with mixed effects.
| stat |
Approximate Bayesian Conditional Copulas | Copula models are flexible tools to represent complex structures of
dependence for multivariate random variables. According to Sklar's theorem
(Sklar, 1959), any d-dimensional absolutely continuous density can be uniquely
represented as the product of the marginal distributions and a copula function
which captures the dependence structure among the vector components. In real
data applications, the interest of the analyses often lies on specific
functionals of the dependence, which quantify aspects of it in a few numerical
values. A broad literature exists on such functionals, however extensions to
include covariates are still limited. This is mainly due to the lack of
unbiased estimators of the copula function, especially when one does not have
enough information to select the copula model. Recent advances in computational
methodologies and algorithms have allowed inference in the presence of
complicated likelihood functions, especially in the Bayesian approach, whose
methods, despite being computationally intensive, allow us to better evaluate
the uncertainty of the estimates. In this work, we present several Bayesian
methods to approximate the posterior distribution of functionals of the
dependence, using nonparametric models which avoid the selection of the copula
function. These methods are compared in simulation studies and in two realistic
applications, from civil engineering and astrophysics.
| stat |
Functional Regression on Manifold with Contamination | We propose a new method for functional nonparametric regression with a
predictor that resides on a finite-dimensional manifold but is only observable
in an infinite-dimensional space. Contamination of the predictor due to
discrete/noisy measurements is also accounted for. By using functional local
linear manifold smoothing, the proposed estimator enjoys a polynomial rate of
convergence that adapts to the intrinsic manifold dimension and the
contamination level. This is in contrast to the logarithmic convergence rate in
the literature of functional nonparametric regression. We also observe a phase
transition phenomenon regarding the interplay of the manifold dimension and the
contamination level. We demonstrate that the proposed method has favorable
numerical performance relative to commonly used methods via simulated and real
data examples.
| stat |
c-lasso -- a Python package for constrained sparse and robust regression
and classification | We introduce c-lasso, a Python package that enables sparse and robust linear
regression and classification with linear equality constraints. The underlying
statistical forward model is assumed to be of the following form: \[ y = X
\beta + \sigma \epsilon \qquad \textrm{subject to} \qquad C\beta=0 \] Here, $X
\in \mathbb{R}^{n\times d}$is a given design matrix and the vector $y \in
\mathbb{R}^{n}$ is a continuous or binary response vector. The matrix $C$ is a
general constraint matrix. The vector $\beta \in \mathbb{R}^{d}$ contains the
unknown coefficients and $\sigma$ an unknown scale. Prominent use cases are
(sparse) log-contrast regression with compositional data $X$, requiring the
constraint $1_d^T \beta = 0$ (Aitchion and Bacon-Shone 1984) and the
Generalized Lasso which is a special case of the described problem (see, e.g,
(James, Paulson, and Rusmevichientong 2020), Example 3). The c-lasso package
provides estimators for inferring unknown coefficients and scale (i.e.,
perspective M-estimators (Combettes and M\"uller 2020a)) of the form \[
\min_{\beta \in \mathbb{R}^d, \sigma \in \mathbb{R}_{0}} f\left(X\beta -
y,{\sigma} \right) + \lambda \left\lVert \beta\right\rVert_1 \qquad
\textrm{subject to} \qquad C\beta = 0 \] for several convex loss functions
$f(\cdot,\cdot)$. This includes the constrained Lasso, the constrained scaled
Lasso, and sparse Huber M-estimators with linear equality constraints.
| stat |
RaSE: Random Subspace Ensemble Classification | We propose a flexible ensemble classification framework, Random Subspace
Ensemble (RaSE), for sparse classification. In the RaSE algorithm, we aggregate
many weak learners, where each weak learner is a base classifier trained in a
subspace optimally selected from a collection of random subspaces. To conduct
subspace selection, we propose a new criterion, ratio information criterion
(RIC), based on weighted Kullback-Leibler divergence. The theoretical analysis
includes the risk and Monte-Carlo variance of RaSE classifier, establishing the
screening consistency and weak consistency of RIC, and providing an upper bound
for the misclassification rate of RaSE classifier. In addition, we show that in
a high-dimensional framework, the number of random subspaces needs to be very
large to guarantee that a subspace covering signals is selected. Therefore, we
propose an iterative version of RaSE algorithm and prove that under some
specific conditions, a smaller number of generated random subspaces are needed
to find a desirable subspace through iteration. An array of simulations under
various models and real-data applications demonstrate the effectiveness and
robustness of the RaSE classifier and its iterative version in terms of low
misclassification rate and accurate feature ranking. The RaSE algorithm is
implemented in the R package RaSEn on CRAN.
| stat |
A novel framework for joint sparse clustering and alignment of
functional data | We propose a novel framework for sparse functional clustering that also
embeds an alignment step. Sparse functional clustering means finding a grouping
structure while jointly detecting the parts of the curves' domains where their
grouping structure shows the most. Misalignment is a well-known issue in
functional data analysis, that can heavily affect functional clustering results
if not properly handled. Therefore, we develop a sparse functional clustering
procedure that accounts for the possible curve misalignment: the coherence of
the functional measure used in the clustering step to the class where the
warping functions are chosen is ensured, and the well-posedness of the sparse
clustering problem is proved. A possible implementing algorithm is also
proposed, that jointly performs all these tasks: clustering, alignment, and
domain selection. The method is tested on simulated data in various realistic
situations, and its application to the Berkeley Growth Study data and to the
AneuRisk65 data set is discussed.
| stat |
The Variational Bayesian Inference for Network Autoregression Models | We develop a variational Bayesian (VB) approach for estimating large-scale
dynamic network models in the network autoregression framework. The VB approach
allows for the automatic identification of the dynamic structure of such a
model and obtains a direct approximation of the posterior density. Compared to
Markov Chain Monte Carlo (MCMC) based sampling approaches, the VB approach
achieves enhanced computational efficiency without sacrificing estimation
accuracy. In the simulation study conducted here, the proposed VB approach
detects various types of proper active structures for dynamic network models.
Compared to the alternative approach, the proposed method achieves similar or
better accuracy, and its computational time is halved. In a real data analysis
scenario of day-ahead natural gas flow prediction in the German gas
transmission network with 51 nodes between October 2013 and September 2015, the
VB approach delivers promising forecasting accuracy along with clearly detected
structures in terms of dynamic dependence.
| stat |
Accumulation Bias in Meta-Analysis: The Need to Consider Time in Error
Control | Studies accumulate over time and meta-analyses are mainly retrospective.
These two characteristics introduce dependencies between the analysis time, at
which a series of studies is up for meta-analysis, and results within the
series. Dependencies introduce bias --- Accumulation Bias --- and invalidate
the sampling distribution assumed for p-value tests, thus inflating type-I
errors. But dependencies are also inevitable, since for science to accumulate
efficiently, new research needs to be informed by past results. Here, we
investigate various ways in which time influences error control in
meta-analysis testing. We introduce an Accumulation Bias Framework that allows
us to model a wide variety of practically occurring dependencies, including
study series accumulation, meta-analysis timing, and approaches to multiple
testing in living systematic reviews. The strength of this framework is that it
shows how all dependencies affect p-value-based tests in a similar manner. This
leads to two main conclusions. First, Accumulation Bias is inevitable, and even
if it can be approximated and accounted for, no valid p-value tests can be
constructed. Second, tests based on likelihood ratios withstand Accumulation
Bias: they provide bounds on error probabilities that remain valid despite the
bias. We leave the reader with a choice between two proposals to consider time
in error control: either treat individual (primary) studies and meta-analyses
as two separate worlds --- each with their own timing --- or integrate
individual studies in the meta-analysis world. Taking up likelihood ratios in
either approach allows for valid tests that relate well to the accumulating
nature of scientific knowledge. Likelihood ratios can be interpreted as betting
profits, earned in previous studies and invested in new ones, while the
meta-analyst is allowed to cash out at any time and advise against future
studies.
| stat |
Automatic Bayesian Density Analysis | Making sense of a dataset in an automatic and unsupervised fashion is a
challenging problem in statistics and AI. Classical approaches for {exploratory
data analysis} are usually not flexible enough to deal with the uncertainty
inherent to real-world data: they are often restricted to fixed latent
interaction models and homogeneous likelihoods; they are sensitive to missing,
corrupt and anomalous data; moreover, their expressiveness generally comes at
the price of intractable inference. As a result, supervision from statisticians
is usually needed to find the right model for the data. However, since domain
experts are not necessarily also experts in statistics, we propose Automatic
Bayesian Density Analysis (ABDA) to make exploratory data analysis accessible
at large. Specifically, ABDA allows for automatic and efficient missing value
estimation, statistical data type and likelihood discovery, anomaly detection
and dependency structure mining, on top of providing accurate density
estimation. Extensive empirical evidence shows that ABDA is a suitable tool for
automatic exploratory analysis of mixed continuous and discrete tabular data.
| stat |
An MCMC Method to Sample from Lattice Distributions | We introduce a Markov Chain Monte Carlo (MCMC) algorithm to generate samples
from probability distributions supported on a $d$-dimensional lattice $\Lambda
= \mathbf{B}\mathbb{Z}^d$, where $\mathbf{B}$ is a full-rank matrix.
Specifically, we consider lattice distributions $P_\Lambda$ in which the
probability at a lattice point is proportional to a given probability density
function, $f$, evaluated at that point. To generate samples from $P_\Lambda$,
it suffices to draw samples from a pull-back measure $P_{\mathbb{Z}^d}$ defined
on the integer lattice. The probability of an integer lattice point under
$P_{\mathbb{Z}^d}$ is proportional to the density function $\pi =
|\det(\mathbf{B})|f\circ \mathbf{B}$. The algorithm we present in this paper
for sampling from $P_{\mathbb{Z}^d}$ is based on the Metropolis-Hastings
framework. In particular, we use $\pi$ as the proposal distribution and
calculate the Metropolis-Hastings acceptance ratio for a well-chosen target
distribution. We can use any method, denoted by ALG, that ideally draws samples
from the probability density $\pi$, to generate a proposed state. The target
distribution is a piecewise sigmoidal distribution, chosen such that the
coordinate-wise rounding of a sample drawn from the target distribution gives a
sample from $P_{\mathbb{Z}^d}$. When ALG is ideal, we show that our algorithm
is uniformly ergodic if $-\log(\pi)$ satisfies a gradient Lipschitz condition.
| stat |
A geostatistical two field model that combines point observations and
nested areal observations, and quantifies long-term spatial variability -- A
case study of annual runoff predictions in the Voss area | In this study, annual runoff is estimated by using a Bayesian geostatistical
model for interpolation of hydrological data of different spatial support. That
is, streamflow observations from catchments (areal data), and precipitation and
evaporation data (point data). The model contains one climatic spatial effect
that is common for all years under study, and one year specific spatial effect.
Hence, the framework enables a quantification of the spatial variability that
is due to long-term weather patterns and processes. This can contribute to a
better understanding of biases and uncertainties in environmental modeling. By
using integrated nested Laplace approximations (INLA) and the stochastic
partial differential equation approach (SPDE) to spatial modeling, the two
field model is computationally feasible and fast. The suggested model is tested
by predicting 10 years of annual runoff around Voss in Norway and through a
simulation study. We find that on average we benefit from combining point and
areal data compared to using only one of the data types, and that the
interaction between nested areal data and point data gives a spatial model that
takes us beyond smoothing. Another finding is that when climatic effects
dominate over annual effects, systematic under- and overestimation of runoff
over time can be expected. On the other hand, a dominating climatic spatial
effect implies that short records of runoff from an otherwise ungauged
catchment can lead to large improvements in the predictability of runoff.
| stat |
Feature space approximation for kernel-based supervised learning | We propose a method for the approximation of high- or even
infinite-dimensional feature vectors, which play an important role in
supervised learning. The goal is to reduce the size of the training data,
resulting in lower storage consumption and computational complexity.
Furthermore, the method can be regarded as a regularization technique, which
improves the generalizability of learned target functions. We demonstrate
significant improvements in comparison to the computation of data-driven
predictions involving the full training data set. The method is applied to
classification and regression problems from different application areas such as
image recognition, system identification, and oceanographic time series
analysis.
| stat |
Latent Distance Estimation for Random Geometric Graphs | Random geometric graphs are a popular choice for a latent points generative
model for networks. Their definition is based on a sample of $n$ points
$X_1,X_2,\cdots,X_n$ on the Euclidean sphere~$\mathbb{S}^{d-1}$ which
represents the latent positions of nodes of the network. The connection
probabilities between the nodes are determined by an unknown function (referred
to as the "link" function) evaluated at the distance between the latent points.
We introduce a spectral estimator of the pairwise distance between latent
points and we prove that its rate of convergence is the same as the
nonparametric estimation of a function on $\mathbb{S}^{d-1}$, up to a
logarithmic factor. In addition, we provide an efficient spectral algorithm to
compute this estimator without any knowledge on the nonparametric link
function. As a byproduct, our method can also consistently estimate the
dimension $d$ of the latent space.
| stat |
Robust approximate Bayesian inference | We discuss an approach for deriving robust posterior distributions from
$M$-estimating functions using Approximate Bayesian Computation (ABC) methods.
In particular, we use $M$-estimating functions to construct suitable summary
statistics in ABC algorithms. The theoretical properties of the robust
posterior distributions are discussed. Special attention is given to the
application of the method to linear mixed models. Simulation results and an
application to a clinical study demonstrate the usefulness of the method. An R
implementation is also provided in the robustBLME package.
| stat |
Compressed Sensing Using Binary Matrices of Nearly Optimal Dimensions | In this paper, we study the problem of compressed sensing using binary
measurement matrices and $\ell_1$-norm minimization (basis pursuit) as the
recovery algorithm. We derive new upper and lower bounds on the number of
measurements to achieve robust sparse recovery with binary matrices. We
establish sufficient conditions for a column-regular binary matrix to satisfy
the robust null space property (RNSP) and show that the associated sufficient
conditions % sparsity bounds for robust sparse recovery obtained using the RNSP
are better by a factor of $(3 \sqrt{3})/2 \approx 2.6$ compared to the
sufficient conditions obtained using the restricted isometry property (RIP).
Next we derive universal \textit{lower} bounds on the number of measurements
that any binary matrix needs to have in order to satisfy the weaker sufficient
condition based on the RNSP and show that bipartite graphs of girth six are
optimal. Then we display two classes of binary matrices, namely parity check
matrices of array codes and Euler squares, which have girth six and are nearly
optimal in the sense of almost satisfying the lower bound. In principle,
randomly generated Gaussian measurement matrices are "order-optimal". So we
compare the phase transition behavior of the basis pursuit formulation using
binary array codes and Gaussian matrices and show that (i) there is essentially
no difference between the phase transition boundaries in the two cases and (ii)
the CPU time of basis pursuit with binary matrices is hundreds of times faster
than with Gaussian matrices and the storage requirements are less. Therefore it
is suggested that binary matrices are a viable alternative to Gaussian matrices
for compressed sensing using basis pursuit. \end{abstract}
| stat |
A Robust Time Series Model with Outliers and Missing Entries | This paper studies the problem of robustly learning the correlation function
for a univariate time series with the presence of noise, outliers and missing
entries. The outliers or anomalies considered here are sparse and rare events
that deviate from normality which is depicted by a correlation function and an
uncertainty condition. This general formulation is applied to univariate time
series of event counts (or non-negative time series) where the correlation is a
log-linear function with the uncertainty condition following the Poisson
distribution. Approximations to the sparsity constraint, such as $\ell^r, 0<
r\le 1$, are used to obtain robustness in the presence of outliers. The
$\ell^r$ constraint is also applied to the correlation function to reduce the
number of active coefficients. This task also helps bypassing the model
selection procedure. Simulated results are presented to validate the model.
| stat |
On Construction and Estimation of Stationary Mixture Transition
Distribution Models | Mixture transition distribution time series models build high-order
dependence through a weighted combination of first-order transition densities
for each one of a specified number of lags. We present a framework to construct
stationary transition mixture distribution models that extend beyond linear,
Gaussian dynamics. We study conditions for first-order strict stationarity
which allow for different constructions with either continuous or discrete
families for the first-order transition densities given a pre-specified family
for the marginal density, and with general forms for the resulting conditional
expectations. Inference and prediction are developed under the Bayesian
framework with particular emphasis on flexible, structured priors for the
mixture weights. Model properties are investigated both analytically and
through synthetic data examples. Finally, Poisson and Lomax examples are
illustrated through real data applications.
| stat |
Variational Auto-Regressive Gaussian Processes for Continual Learning | This paper proposes Variational Auto-Regressive Gaussian Process (VAR-GP), a
principled Bayesian updating mechanism to incorporate new data for sequential
tasks in the context of continual learning. It relies on a novel
auto-regressive characterization of the variational distribution and inference
is made scalable using sparse inducing point approximations. Experiments on
standard continual learning benchmarks demonstrate the ability of VAR-GPs to
perform well at new tasks without compromising performance on old ones,
yielding competitive results to state-of-the-art methods. In addition, we
qualitatively show how VAR-GP improves the predictive entropy estimates as we
train on new tasks. Further, we conduct a thorough ablation study to verify the
effectiveness of inferential choices.
| stat |
Pushing the Limit: A Hybrid Parallel Implementation of the
Multi-resolution Approximation for Massive Data | The multi-resolution approximation (MRA) of Gaussian processes was recently
proposed to conduct likelihood-based inference for massive spatial data sets.
An advantage of the methodology is that it can be parallelized. We implemented
the MRA in C++ for both serial and parallel versions. In the parallel
implementation, we use a hybrid parallelism that employs both distributed and
shared memory computing for communications between and within nodes by using
the Message Passing Interface (MPI) and OpenMP, respectively. The performance
of the serial code is compared between the C++ and MATLAB implementations over
a small data set on a personal laptop. The C++ parallel program is further
carefully studied under different configurations by applications to data sets
from around a tenth of a million to 47 million observations. We show the
practicality of this implementation by demonstrating that we can get quick
inference for massive real-world data sets. The serial and parallel C++ code
can be found at https://github.com/hhuang90.
| stat |
Manifold Markov chain Monte Carlo methods for Bayesian inference in a
wide class of diffusion models | Bayesian inference for nonlinear diffusions, observed at discrete times, is a
challenging task that has prompted the development of a number of algorithms,
mainly within the computational statistics community. We propose a new
direction, and accompanying methodology, borrowing ideas from statistical
physics and computational chemistry, for inferring the posterior distribution
of latent diffusion paths and model parameters, given observations of the
process. Joint configurations of the underlying process noise and of
parameters, mapping onto diffusion paths consistent with observations, form an
implicitly defined manifold. Then, by making use of a constrained Hamiltonian
Monte Carlo algorithm on the embedded manifold, we are able to perform
computationally efficient inference for an extensive class of discretely
observed diffusion models. Critically, in contrast with other approaches
proposed in the literature, our methodology is highly automated, requiring
minimal user intervention and applying alike in a range of settings, including:
elliptic or hypo-elliptic systems; observations with or without noise; linear
or non-linear observation operators. Exploiting Markovianity, we propose a
variant of the method with complexity that scales linearly in the resolution of
path discretisation and the number of observation times.
| stat |
Semi-supervised Anomaly Detection on Attributed Graphs | We propose a simple yet effective method for detecting anomalous instances on
an attribute graph with label information of a small number of instances.
Although with standard anomaly detection methods it is usually assumed that
instances are independent and identically distributed, in many real-world
applications, instances are often explicitly connected with each other,
resulting in so-called attributed graphs. The proposed method embeds nodes
(instances) on the attributed graph in the latent space by taking into account
their attributes as well as the graph structure based on graph convolutional
networks (GCNs). To learn node embeddings specialized for anomaly detection, in
which there is a class imbalance due to the rarity of anomalies, the parameters
of a GCN are trained to minimize the volume of a hypersphere that encloses the
node embeddings of normal instances while embedding anomalous ones outside the
hypersphere. This enables us to detect anomalies by simply calculating the
distances between the node embeddings and hypersphere center. The proposed
method can effectively propagate label information on a small amount of nodes
to unlabeled ones by taking into account the node's attributes, graph
structure, and class imbalance. In experiments with five real-world attributed
graph datasets, we demonstrate that the proposed method achieves better
performance than various existing anomaly detection methods.
| stat |
The impact of modelling choices on modelling outcomes: a spatio-temporal
study of the association between COVID-19 spread and environmental conditions
in Catalonia (Spain) | The choices that researchers make while conducting a statistical analysis
usually have a notable impact on the results. This fact has become evident in
the ongoing research of the association between the environment and the
evolution of the COVID-19 pandemic, in light of the hundreds of contradictory
studies that have already been published on this issue in just a few months. In
this paper, a COVID-19 dataset containing the number of daily cases registered
in the regions of Catalonia (Spain) since the start of the pandemic is
analysed. Specifically, the possible effect of several environmental variables
(solar exposure, mean temperature, and wind speed) on the number of cases is
assessed. Thus, the first objective of the paper is to show how the choice of a
certain type of statistical model to conduct the analysis can have a severe
impact on the associations that are inferred between the covariates and the
response variable. Secondly, it is shown how the use of spatio-temporal models
accounting for the nature of the data allows understanding the evolution of the
pandemic in space and time.
| stat |
Multiple Imputation for Nonignorable Item Nonresponse in Complex Surveys
Using Auxiliary Margin | We outline a framework for multiple imputation of nonignorable item
nonresponse when the marginal distributions of some of the variables with
missing values are known. In particular, our framework ensures that (i) the
completed datasets result in design-based estimates of totals that are
plausible, given the margins, and (ii) the completed datasets maintain
associations across variables as posited in the imputation models. To do so, we
propose an additive nonignorable model for nonresponse, coupled with a
rejection sampling step. The rejection sampling step favors completed datasets
that result in design-based estimates that are plausible given the known
margins. We illustrate the framework using simulations with stratified
sampling.
| stat |
Asymptotic Behavior of Adversarial Training in Binary Classification | It has been consistently reported that many machine learning models are
susceptible to adversarial attacks i.e., small additive adversarial
perturbations applied to data points can cause misclassification. Adversarial
training using empirical risk minimization is considered to be the
state-of-the-art method for defense against adversarial attacks. Despite being
successful in practice, several problems in understanding generalization
performance of adversarial training remain open. In this paper, we derive
precise theoretical predictions for the performance of adversarial training in
binary classification. We consider the high-dimensional regime where the
dimension of data grows with the size of the training data-set at a constant
ratio. Our results provide exact asymptotics for standard and adversarial
errors of the estimators obtained by adversarial training with $\ell_q$-norm
bounded perturbations ($q \ge 1$) for both discriminative binary models and
generative Gaussian mixture models. Furthermore, we use these sharp predictions
to uncover several intriguing observations on the role of various parameters
including the over-parameterization ratio, the data model, and the attack
budget on the adversarial and standard errors.
| stat |
True and false discoveries with independent e-values | In this note we use e-values (a non-Bayesian version of Bayes factors) in the
context of multiple hypothesis testing assuming that the base tests produce
independent e-values. Our simulation studies and theoretical considerations
suggest that, under this assumption, our new algorithms are superior to the
known algorithms using independent p-values and to our recent algorithms using
e-values that are not necessarily independent.
| stat |
Characterizing Implicit Bias in Terms of Optimization Geometry | We study the implicit bias of generic optimization methods, such as mirror
descent, natural gradient descent, and steepest descent with respect to
different potentials and norms, when optimizing underdetermined linear
regression or separable linear classification problems. We explore the question
of whether the specific global minimum (among the many possible global minima)
reached by an algorithm can be characterized in terms of the potential or norm
of the optimization geometry, and independently of hyperparameter choices such
as step-size and momentum.
| stat |
Towards Efficient Connected and Automated Driving System via Multi-agent
Graph Reinforcement Learning | Connected and automated vehicles (CAVs) have attracted more and more
attention recently. The fast actuation time allows them having the potential to
promote the efficiency and safety of the whole transportation system. Due to
technical challenges, there will be a proportion of vehicles that can be
equipped with automation while other vehicles are without automation. Instead
of learning a reliable behavior for ego automated vehicle, we focus on how to
improve the outcomes of the total transportation system by allowing each
automated vehicle to learn cooperation with each other and regulate
human-driven traffic flow. One of state of the art method is using
reinforcement learning to learn intelligent decision making policy. However,
direct reinforcement learning framework cannot improve the performance of the
whole system. In this article, we demonstrate that considering the problem in
multi-agent setting with shared policy can help achieve better system
performance than non-shared policy in single-agent setting. Furthermore, we
find that utilization of attention mechanism on interaction features can
capture the interplay between each agent in order to boost cooperation. To the
best of our knowledge, while previous automated driving studies mainly focus on
enhancing individual's driving performance, this work serves as a starting
point for research on system-level multi-agent cooperation performance using
graph information sharing. We conduct extensive experiments in car-following
and unsignalized intersection settings. The results demonstrate that CAVs
controlled by our method can achieve the best performance against several state
of the art baselines.
| stat |
Causes of Misleading Statistics and Research Results Irreproducibility:
A Concise Review | Bad statistics make research papers unreproducible and misleading. For the
most part, the reasons for such misusage of numerical data have been found and
addressed years ago by experts and proper practical solutions have been
presented instead. Yet, we still see numerous instances of statistical
fallacies in modern researches which without a doubt play a significant role in
the research reproducibility crisis. In this paper, we review different bad
practices that impact the research process from its beginning to its very end.
Additionally, we briefly propose open science as a universal methodology that
can facilitate the entire research life cycle.
| stat |
Random Forests for dependent data | Random forest (RF) is one of the most popular methods for estimating
regression functions. The local nature of the RF algorithm, based on intra-node
means and variances, is ideal when errors are i.i.d. For dependent error
processes like time series and spatial settings where data in all the nodes
will be correlated, operating locally ignores this dependence. Also, RF will
involve resampling of correlated data, violating the principles of bootstrap.
Theoretically, consistency of RF has been established for i.i.d. errors, but
little is known about the case of dependent errors.
We propose RF-GLS, a novel extension of RF for dependent error processes in
the same way Generalized Least Squares (GLS) fundamentally extends Ordinary
Least Squares (OLS) for linear models under dependence. The key to this
extension is the equivalent representation of the local decision-making in a
regression tree as a global OLS optimization which is then replaced with a GLS
loss to create a GLS-style regression tree. This also synergistically addresses
the resampling issue, as the use of GLS loss amounts to resampling uncorrelated
contrasts (pre-whitened data) instead of the correlated data. For spatial
settings, RF-GLS can be used in conjunction with Gaussian Process correlated
errors to generate kriging predictions at new locations. RF becomes a special
case of RF-GLS with an identity working covariance matrix.
We establish consistency of RF-GLS under beta- (absolutely regular) mixing
error processes and show that this general result subsumes important cases like
autoregressive time series and spatial Matern Gaussian Processes. As a
byproduct, we also establish consistency of RF for beta-mixing processes, which
to our knowledge, is the first such result for RF under dependence.
We empirically demonstrate the improvement achieved by RF-GLS over RF for
both estimation and prediction under dependence.
| stat |
Powerful Knockoffs via Minimizing Reconstructability | Model-X knockoffs allows analysts to perform feature selection using almost
any machine learning algorithm while still provably controlling the expected
proportion of false discoveries. To apply model-X knockoffs, one must construct
synthetic variables, called knockoffs, which effectively act as controls during
feature selection. The gold standard for constructing knockoffs has been to
minimize the mean absolute correlation (MAC) between features and their
knockoffs, but, surprisingly, we prove this procedure can be powerless in
extremely easy settings, including Gaussian linear models with correlated
exchangeable features. The key problem is that minimizing the MAC creates
strong joint dependencies between the features and knockoffs, which allow
machine learning algorithms to partially or fully reconstruct the effect of the
features on the response using the knockoffs. To improve the power of
knockoffs, we propose generating knockoffs which minimize the
reconstructability (MRC) of the features, and we demonstrate our proposal for
Gaussian features by showing it is computationally efficient, robust, and
powerful. We also prove that certain MRC knockoffs minimize a natural
definition of estimation error in Gaussian linear models. Furthermore, in an
extensive set of simulations, we find many settings with correlated features in
which MRC knockoffs dramatically outperform MAC-minimizing knockoffs and no
settings in which MAC-minimizing knockoffs outperform MRC knockoffs by more
than a very slight margin. We implement our methods and a host of others from
the knockoffs literature in a new open source python package knockpy.
| stat |
Learning and Inference in Imaginary Noise Models | Inspired by recent developments in learning smoothed densities with empirical
Bayes, we study variational autoencoders with a decoder that is tailored for
the random variable $Y=X+N(0,\sigma^2 I_d)$. A notion of smoothed variational
inference emerges where the smoothing is implicitly enforced by the noise model
of the decoder; "implicit", since during training the encoder only sees clean
samples. This is the concept of imaginary noise model, where the noise model
dictates the functional form of the variational lower bound
$\mathcal{L}(\sigma)$, but the noisy data are never seen during learning. The
model is named $\sigma$-VAE. We prove that all $\sigma$-VAEs are equivalent to
each other via a simple $\beta$-VAE expansion: $\mathcal{L}(\sigma_2) \equiv
\mathcal{L}(\sigma_1,\beta)$, where $\beta=\sigma_2^2/\sigma_1^2$. We prove a
similar result for the Laplace distribution in exponential families.
Empirically, we report an intriguing power law $\mathcal{D}_{\rm KL} \sim
\sigma^{-\nu}$ for the learned models and we study the inference in the
$\sigma$-VAE for unseen noisy data. The experiments were performed on MNIST,
where we show that quite remarkably the model can make reasonable inferences on
extremely noisy samples even though it has not seen any during training. The
vanilla VAE completely breaks down in this regime. We finish with a hypothesis
(the XYZ hypothesis) on the findings here.
| stat |
Random family method: Confirming inter-generational relations by
restricted re-sampling | Randomness is one of the important key concepts of statistics. In
epidemiology or medical science, we investigate our hypotheses and interpret
results through this statistical randomness. We hypothesized by imposing some
conditions to this randomness, interpretation of our result may be changed. In
this article, we introduced the restricted re-sampling method to confirm
inter-generational relations and presented an example.
| stat |
Estimating the Design Operating Characteristics in Clinical Trials with
the Ordinal Scale Disease Progression Endpoint | Bayesian adaptive designs have gained popularity in all phases of clinical
trials in the recent years. The COVID-19 pandemic, however, has brought these
designs to the centre stage. The need for establishing evidence for the
effectiveness of vaccines, therapeutic treatments and policies that could
resolve or control the crisis has resulted in development of efficient designs
for clinical trials that can be concluded with smaller sample sizes in a
shorter time period. Design of Bayesian adaptive trials, however, requires
extensive simulation studies that is considered a disadvantage in
time-sensitive settings such as the pandemic. In this paper, we propose a set
of methods for efficient estimation and uncertainty quantification for the
design operating characteristics of Bayesian adaptive trials. The proposed
approach is tailored to address design of clinical trials with the ordinal
disease progression scale endpoint but can be used generally in the clinical
trials context where design operating characteristics cannot be obtained
analytically.
| stat |
Hierarchical Probabilistic Model for Blind Source Separation via
Legendre Transformation | We present a novel blind source separation (BSS) method, called information
geometric blind source separation (IGBSS). Our formulation is based on the
information geometric log-linear model equipped with a hierarchically
structured sample space, which has theoretical guarantees to uniquely recover a
set of source signals by minimizing the KL divergence from a set of mixed
signals. Source signals, received signals, and mixing matrices are realized as
different layers in our hierarchical sample space. Our empirical results have
demonstrated on images and time series data that our approach is superior to
current state-of-the-art techniques and is able to separate signals with
complex interactions.
| stat |
Parameter Redundancy and the Existence of Maximum Likelihood Estimates
in Log-linear Models | Log-linear models are typically fitted to contingency table data to describe
and identify the relationship between different categorical variables. However,
the data may include observed zero cell entries. The presence of zero cell
entries can have an adverse effect on the estimability of parameters, due to
parameter redundancy. We describe a general approach for determining whether a
given log-linear model is parameter redundant for a pattern of observed zeros
in the table, prior to fitting the model to the data. We derive the estimable
parameters or functions of parameters and also explain how to reduce the
unidentifiable model to an identifiable one. Parameter redundant models have a
flat ridge in their likelihood function. We further explain when this ridge
imposes some additional parameter constraints on the model, which can lead to
obtaining unique maximum likelihood estimates for parameters that otherwise
would not have been estimable. In contrast to other frameworks, the proposed
novel approach informs on those constraints, elucidating the model that is
actually being fitted.
| stat |
A note on tools for prediction under uncertainty and identifiability of
SIR-like dynamical systems for epidemiology | We provide an overview of the methods that can be used for prediction under
uncertainty and data fitting of dynamical systems, and of the fundamental
challenges that arise in this context. The focus is on SIR-like models, that
are being commonly used when attempting to predict the trend of the COVID-19
pandemic. In particular, we raise a warning flag about identifiability of the
parameters of SIR-like models; often, it might be hard to infer the correct
values of the parameters from data, even for very simple models, making it
non-trivial to use these models for meaningful predictions. Most of the points
that we touch upon are actually generally valid for inverse problems in more
general setups.
| stat |
Pseudo Bayesian Mixed Models under Informative Sampling | When random effects are correlated with sample design variables, the usual
approach of employing individual survey weights (constructed to be inversely
proportional to the unit survey inclusion probabilities) to form a
pseudo-likelihood no longer produces asymptotically unbiased inference. We
construct a weight-exponentiated formulation for the random effects
distribution that achieves unbiased inference for generating hyperparameters of
the random effects. We contrast our approach with frequentist methods that rely
on numerical integration to reveal that only the Bayesian method achieves both
unbiased estimation with respect to the sampling design distribution and
consistency with respect to the population generating distribution. Our
simulations and real data example for a survey of business establishments
demonstrate the utility of our approach across different modeling formulations
and sampling designs. This work serves as a capstone for recent developmental
efforts that combine traditional survey estimation approaches with the Bayesian
modeling paradigm and provides a bridge across the two rich but disparate
sub-fields.
| stat |
The effect of measurement error on clustering algorithms | Clustering consists of a popular set of techniques used to separate data into
interesting groups for further analysis. Many data sources on which clustering
is performed are well-known to contain random and systematic measurement
errors. Such errors may adversely affect clustering. While several techniques
have been developed to deal with this problem, little is known about the
effectiveness of these solutions. Moreover, no work to-date has examined the
effect of systematic errors on clustering solutions.
In this paper, we perform a Monte Carlo study to investigate the sensitivity
of two common clustering algorithms, GMMs with merging and DBSCAN, to random
and systematic error. We find that measurement error is particularly
problematic when it is systematic and when it affects all variables in the
dataset. For the conditions considered here, we also find that the
partition-based GMM with merged components is less sensitive to measurement
error than the density-based DBSCAN procedure.
| stat |
A nonparametric empirical Bayes approach to covariance matrix estimation | We propose an empirical Bayes method to estimate high-dimensional covariance
matrices. Our procedure centers on vectorizing the covariance matrix and
treating matrix estimation as a vector estimation problem. Drawing from the
compound decision theory literature, we introduce a new class of decision rules
that generalizes several existing procedures. We then use a nonparametric
empirical Bayes g-modeling approach to estimate the oracle optimal rule in that
class. This allows us to let the data itself determine how best to shrink the
estimator, rather than shrinking in a pre-determined direction such as toward a
diagonal matrix. Simulation results and a gene expression network analysis
shows that our approach can outperform a number of state-of-the-art proposals
in a wide range of settings, sometimes substantially.
| stat |
Causality and Robust Optimization | A decision-maker must consider cofounding bias when attempting to apply
machine learning prediction, and, while feature selection is widely recognized
as important process in data-analysis, it could cause cofounding bias. A causal
Bayesian network is a standard tool for describing causal relationships, and if
relationships are known, then adjustment criteria can determine with which
features cofounding bias disappears. A standard modification would thus utilize
causal discovery algorithms for preventing cofounding bias in feature
selection. Causal discovery algorithms, however, essentially rely on the
faithfulness assumption, which turn out to be easily violated in practical
feature selection settings. In this paper, we propose a meta-algorithm that can
remedy existing feature selection algorithms in terms of cofounding bias. Our
algorithm is induced from a novel adjustment criterion that requires rather
than faithfulness, an assumption which can be induced from another well-known
assumption of the causal sufficiency. We further prove that the features added
through our modification convert cofounding bias into prediction variance. With
the aid of existing robust optimization technologies that regularize risky
strategies with high variance, then, we are able to successfully improve the
throughput performance of decision-making optimization, as is shown in our
experimental results.
| stat |
Estimation of future discretionary benefits in traditional life
insurance | In the context of traditional life insurance, the future discretionary
benefits ($FDB$), which are a central item for Solvency~II reporting, are
generally calculated by computationally expensive Monte Carlo algorithms. We
derive analytic formulas for lower and upper bounds for the $FDB$. This yields
an estimation interval for the $FDB$, and the average of lower and upper bound
is a simple estimator. These formulae are designed for real world applications,
and we compare the results to publicly available reporting data.
| stat |
Backward Simulation of Multivariate Mixed Poisson Processes | The Backward Simulation (BS) approach was developed to generate, simply and
efficiently, sample paths of correlated multivariate Poisson process with
negative correlation coefficients between their components. In this paper, we
extend the BS approach to model multivariate Mixed Poisson processes which have
many important applications in Insurance, Finance, Geophysics and many other
areas of Applied Probability. We also extend the Forward Continuation approach,
introduced in our earlier work, to multivariate Mixed Poisson processes.
| stat |
Multiple kernel learning for integrative consensus clustering of 'omic
datasets | Diverse applications - particularly in tumour subtyping - have demonstrated
the importance of integrative clustering techniques for combining information
from multiple data sources. Cluster-Of-Clusters Analysis (COCA) is one such
approach that has been widely applied in the context of tumour subtyping.
However, the properties of COCA have never been systematically explored, and
its robustness to the inclusion of noisy datasets, or datasets that define
conflicting clustering structures, is unclear. We rigorously benchmark COCA,
and present Kernel Learning Integrative Clustering (KLIC) as an alternative
strategy. KLIC frames the challenge of combining clustering structures as a
multiple kernel learning problem, in which different datasets each provide a
weighted contribution to the final clustering. This allows the contribution of
noisy datasets to be down-weighted relative to more informative datasets. We
compare the performances of KLIC and COCA in a variety of situations through
simulation studies. We also present the output of KLIC and COCA in real data
applications to cancer subtyping and transcriptional module discovery. R
packages "klic" and "coca" are available on the Comprehensive R Archive
Network.
| stat |
Geodesics in fibered latent spaces: A geometric approach to learning
correspondences between conditions | This work introduces a geometric framework and a novel network architecture
for creating correspondences between samples of different conditions. Under
this formalism, the latent space is a fiber bundle stratified into a base space
encoding conditions, and a fiber space encoding the variations within
conditions. Furthermore, this latent space is endowed with a natural pull-back
metric. The correspondences between conditions are obtained by minimizing an
energy functional, resulting in diffeomorphism flows between fibers.
We illustrate this approach using MNIST and Olivetti and benchmark its
performances on the task of batch correction, which is the problem of
integrating multiple biological datasets together.
| stat |
Leaders, Followers, and Community Detection | Communities in social networks or graphs are sets of well-connected,
overlapping vertices. The effectiveness of a community detection algorithm is
determined by accuracy in finding the ground-truth communities and ability to
scale with the size of the data. In this work, we provide three contributions.
First, we show that a popular measure of accuracy known as the F1 score, which
is between 0 and 1, with 1 being perfect detection, has an information lower
bound is 0.5. We provide a trivial algorithm that produces communities with an
F1 score of 0.5 for any graph! Somewhat surprisingly, we find that popular
algorithms such as modularity optimization, BigClam and CESNA have F1 scores
less than 0.5 for the popular IMDB graph. To rectify this, as the second
contribution we propose a generative model for community formation, the
sequential community graph, which is motivated by the formation of social
networks. Third, motivated by our generative model, we propose the
leader-follower algorithm (LFA). We prove that it recovers all communities for
sequential community graphs by establishing a structural result that sequential
community graphs are chordal. For a large number of popular social networks, it
recovers communities with a much higher F1 score than other popular algorithms.
For the IMDB graph, it obtains an F1 score of 0.81. We also propose a
modification to the LFA called the fast leader-follower algorithm (FLFA) which
in addition to being highly accurate, is also fast, with a scaling that is
almost linear in the network size.
| stat |
Trends in risks of severe events and lengths of stay for COVID-19
hospitalisations in England over the pre-vaccination era: results from the
Public Health England SARI-Watch surveillance scheme | Background: Trends in hospitalised case-fatality risk (HFR), risk of
intensive care unit (ICU) admission and lengths of stay for patients
hospitalised for COVID-19 in England over the pre-vaccination era are unknown.
Methods: Data on hospital and ICU admissions with COVID-19 at 31 NHS trusts
in England were collected by Public Health England's Severe Acute Respiratory
Infections surveillance system and linked to death information. We applied
parametric multi-state mixture models, accounting for censored outcomes and
regressing risks and times between events on month of admission, geography, and
baseline characteristics.
Findings: 20,785 adults were admitted with COVID-19 in 2020. Between March
and June/July/August estimated HFR reduced from 31.9% (95% confidence interval
30.3-33.5%) to 10.9% (9.4-12.7%), then rose steadily from 21.6% (18.4-25.5%) in
September to 25.7% (23.0-29.2%) in December, with steeper increases among older
patients, those with multi-morbidity and outside London/South of England. ICU
admission risk reduced from 13.9% (12.8-15.2%) in March to 6.2% (5.3-7.1%) in
May, rising to a high of 14.2% (11.1-17.2%) in September. Median length of stay
in non-critical care increased during 2020, from 6.6 to 12.3 days for those
dying, and from 6.1 to 9.3 days for those discharged.
Interpretation: Initial improvements in patient outcomes, corresponding to
developments in clinical practice, were not sustained throughout 2020, with HFR
in December approaching the levels seen at the start of the pandemic, whilst
median hospital stays have lengthened. The role of increased transmission, new
variants, case-mix and hospital pressures in increasing COVID-19 severity
requires urgent further investigation.
| stat |
Integrating omics and MRI data with kernel-based tests and CNNs to
identify rare genetic markers for Alzheimer's disease | For precision medicine and personalized treatment, we need to identify
predictive markers of disease. We focus on Alzheimer's disease (AD), where
magnetic resonance imaging scans provide information about the disease status.
By combining imaging with genome sequencing, we aim at identifying rare genetic
markers associated with quantitative traits predicted from convolutional neural
networks (CNNs), which traditionally have been derived manually by experts.
Kernel-based tests are a powerful tool for associating sets of genetic
variants, but how to optimally model rare genetic variants is still an open
research question. We propose a generalized set of kernels that incorporate
prior information from various annotations and multi-omics data. In the
analysis of data from the Alzheimer's Disease Neuroimaging Initiative (ADNI),
we evaluate whether (i) CNNs yield precise and reliable brain traits, and (ii)
the novel kernel-based tests can help to identify loci associated with AD. The
results indicate that CNNs provide a fast, scalable and precise tool to derive
quantitative AD traits and that new kernels integrating domain knowledge can
yield higher power in association tests of very rare variants.
| stat |
Nonlinear Independent Component Analysis for Continuous-Time Signals | We study the classical problem of recovering a multidimensional source
process from observations of nonlinear mixtures of this process. Assuming
statistical independence of the coordinate processes of the source, we show
that this recovery is possible for many popular models of stochastic processes
(up to order and monotone scaling of their coordinates) if the mixture is given
by a sufficiently differentiable, invertible function. Key to our approach is
the combination of tools from stochastic analysis and recent contrastive
learning approaches to nonlinear ICA. This yields a scalable method with widely
applicable theoretical guarantees for which our experiments indicate good
performance.
| stat |
GraKeL: A Graph Kernel Library in Python | The problem of accurately measuring the similarity between graphs is at the
core of many applications in a variety of disciplines. Graph kernels have
recently emerged as a promising approach to this problem. There are now many
kernels, each focusing on different structural aspects of graphs. Here, we
present GraKeL, a library that unifies several graph kernels into a common
framework. The library is written in Python and adheres to the scikit-learn
interface. It is simple to use and can be naturally combined with
scikit-learn's modules to build a complete machine learning pipeline for tasks
such as graph classification and clustering. The code is BSD licensed and is
available at: https://github.com/ysig/GraKeL .
| stat |
Statistical dynamics of social distancing in SARS-CoV-2 as a
differential game | The novel coronavirus SARS-CoV-2 has rapidly emerged as a significant threat
to global public health, in particular because -- as is not uncommon with novel
pathogens -- there is no effective pharmaceutical treatment or prophylaxis to
the viral syndrome it causes. In the absence of such specific treatment
modalities, the mainstay of public health response rests on non-pharmaceutical
interventions (NPIs), such as social distancing. This paper contributes to the
understanding of social distancing against SARS-CoV-2 by quantitatively
analysing the statistical dynamics of disease propagation as a differential
game, and estimating the relative costs of distancing versus not distancing,
identifying marginal utility of distancing based on known population
epidemiological data about SARS-CoV-2 and concluding that unless the costs of
distancing vastly exceed the cost of illness per unit time, social distancing
remains a dominant strategy. These findings can assist in solidly anchoring
public health responses based on social distancing within a quantitative
framework attesting to their effectiveness.
| stat |
Autoregressive Density Modeling with the Gaussian Process Mixture
Transition Distribution | We develop a mixture model for transition density approximation, together
with soft model selection, in the presence of noisy and heterogeneous nonlinear
dynamics. Our model builds on the Gaussian mixture transition distribution
(MTD) model for continuous state spaces, extending component means with
nonlinear functions that are modeled using Gaussian process (GP) priors. The
resulting model flexibly captures nonlinear and heterogeneous lag dependence
when several mixture components are active, identifies low-order nonlinear
dependence while inferring relevant lags when few components are active, and
averages over multiple and competing single-lag models to quantify/propagate
uncertainty. Sparsity-inducing priors on the mixture weights aid in selecting a
subset of active lags. The hierarchical model specification follows conventions
for both GP regression and MTD models, admitting a convenient Gibbs sampling
scheme for posterior inference. We demonstrate properties of the proposed model
with two simulated and two real time series, emphasizing approximation of
lag-dependent transition densities and model selection. In most cases, the
model decisively recovers important features. The proposed model provides a
simple, yet flexible framework that preserves useful and distinguishing
characteristics of the MTD model class.
| stat |
Revisiting Rashomon: A Comment on "The Two Cultures" | Here, I provide some reflections on Prof. Leo Breiman's "The Two Cultures"
paper. I focus specifically on the phenomenon that Breiman dubbed the "Rashomon
Effect", describing the situation in which there are many models that satisfy
predictive accuracy criteria equally well, but process information in the data
in substantially different ways. This phenomenon can make it difficult to draw
conclusions or automate decisions based on a model fit to data. I make
connections to recent work in the Machine Learning literature that explore the
implications of this issue, and note that grappling with it can be a fruitful
area of collaboration between the algorithmic and data modeling cultures.
| stat |
On Misspecification in Prediction Problems and Robustness via Improper
Learning | We study probabilistic prediction games when the underlying model is
misspecified, investigating the consequences of predicting using an incorrect
parametric model. We show that for a broad class of loss functions and
parametric families of distributions, the regret of playing a "proper"
predictor -- one from the putative model class -- relative to the best
predictor in the same model class has lower bound scaling at least as
$\sqrt{\gamma n}$, where $\gamma$ is a measure of the model misspecification to
the true distribution in terms of total variation distance. In contrast, using
an aggregation-based (improper) learner, one can obtain regret $d \log n$ for
any underlying generating distribution, where $d$ is the dimension of the
parameter; we exhibit instances in which this is unimprovable even over the
family of all learners that may play distributions in the convex hull of the
parametric family. These results suggest that simple strategies for aggregating
multiple learners together should be more robust, and several experiments
conform to this hypothesis.
| stat |
Monotonic Nonparametric Dose Response Model | Toxicologists are often concerned with determining the dosage to which an
individual can be exposed with an acceptable risk of adverse effect. These
types of studies have been conducted widely in the past, and many novel
approaches have been developed. Parametric techniques utilizing ANOVA and
nonlinear regression models are well represented in the literature. The biggest
drawback of parametric approaches is the need to specify the correct model.
Recently, there has been an interest in nonparametric approaches to tolerable
dosage estimation. In this work, we focus on the monotonically decreasing dose
response model where the response is a percent to control. This poses two
constraints to the nonparametric approach. The doseresponse function must be
one at control (dose = 0), and the function must always be positive. Here we
propose a Bayesian solution to this problem using a novel class of
nonparametric models. A basis function developed in this research is the Alamri
Monotonic spline (AM-spline). Our approach is illustrated using both simulated
data and an experimental dataset from pesticide related research at the US
Environmental Protection Agency.
| stat |
Latent-state models for precision medicine | Observational longitudinal studies are a common means to study treatment
efficacy and safety in chronic mental illness. In many such studies, treatment
changes may be initiated by either the patient or by their clinician and can
thus vary widely across patients in their timing, number, and type. Indeed, in
the observational longitudinal pathway of the STEP-BD study of bipolar
depression, one of the motivations for this work, no two patients have the same
treatment history even after coarsening clinic visits to a weekly time-scale.
Estimation of an optimal treatment regime using such data is challenging as one
cannot naively pool together patients with the same treatment history, as is
required by methods based on inverse probability weighting, nor is it possible
to apply backwards induction over the decision points, as is done in Q-learning
and its variants. Thus, additional structure is needed to effectively pool
information across patients and within a patient over time. Current scientific
theory for many chronic mental illnesses maintains that a patient's disease
status can be conceptualized as transitioning among a small number of discrete
states. We use this theory to inform the construction of a partially observable
Markov decision process model of patient health trajectories wherein observed
health outcomes are dictated by a patient's latent health state. Using this
model, we derive and evaluate estimators of an optimal treatment regime under
two common paradigms for quantifying long-term patient health. The finite
sample performance of the proposed estimator is demonstrated through a series
of simulation experiments and application to the observational pathway of the
STEP-BD study. We find that the proposed method provides high-quality estimates
of an optimal treatment strategy in settings where existing approaches cannot
be applied without ad hoc modifications.
| stat |
Robust and flexible inference for the covariate-specific ROC curve | Diagnostic tests are of critical importance in health care and medical
research. Motivated by the impact that atypical and outlying test outcomes
might have on the assessment of the discriminatory ability of a diagnostic
test, we develop a flexible and robust model for conducting inference about the
covariate-specific receiver operating characteristic (ROC) curve that
safeguards against outlying test results while also accommodating for possible
nonlinear effects of the covariates. Specifically, we postulate a
location-scale additive regression model for the test outcomes in both the
diseased and nondiseased populations, combining additive cubic B-splines and
M-estimation for the regression function, while the residuals are estimated via
a weighted empirical distribution function. The results of the simulation study
show that our approach successfully recovers the true covariate-specific ROC
curve and corresponding area under the curve on a variety of conceivable test
outcomes contamination scenarios. Our method is applied to a dataset derived
from a prostate cancer study where we seek to assess the ability of the
Prostate Health Index to discriminate between men with and without Gleason 7 or
above prostate cancer, and if and how such discriminatory capacity changes with
age.
| stat |
An introduction to the determination of the probability of a successful
trial: Frequentist and Bayesian approaches | Determination of posterior probability for go-no-go decision and predictive
power are becoming increasingly common for resource optimization in clinical
investigation. There are vast published literature on these topics; however,
the terminologies are not consistently used across the literature. Further,
there is a lack of consolidated presentation of various concepts of the
probability of success. We attempted to fill this gap. This paper first
provides a detailed derivation of these probability of success measures under
the frequentist and Bayesian paradigms in a general setting. Subsequently, we
have presented the analytical formula for these probability of success measures
for continuous, binary, and time-to-event endpoints separately. This paper can
be used as a single point reference to determine the following measures: (a)
the conditional power (CP) based on interim results, (b) the predictive power
of success (PPoS) based on interim results with or without prior distribution,
and (d) the probability of success (PoS) for a prospective trial at the design
stage. We have discussed both clinical success and trial success. This paper's
discussion is mostly based on the normal approximation for prior distribution
and the estimate of the parameter of interest. Besides, predictive power using
the beta prior for the binomial case is also presented. Some examples are given
for illustration. R functions to calculate CP and PPoS are available through
the LongCART package. An R shiny app is also available at
https://ppos.herokuapp.com/.
| stat |
Valid Causal Inference with (Some) Invalid Instruments | Instrumental variable methods provide a powerful approach to estimating
causal effects in the presence of unobserved confounding. But a key challenge
when applying them is the reliance on untestable "exclusion" assumptions that
rule out any relationship between the instrument variable and the response that
is not mediated by the treatment. In this paper, we show how to perform
consistent IV estimation despite violations of the exclusion assumption. In
particular, we show that when one has multiple candidate instruments, only a
majority of these candidates---or, more generally, the modal candidate-response
relationship---needs to be valid to estimate the causal effect. Our approach
uses an estimate of the modal prediction from an ensemble of instrumental
variable estimators. The technique is simple to apply and is "black-box" in the
sense that it may be used with any instrumental variable estimator as long as
the treatment effect is identified for each valid instrument independently. As
such, it is compatible with recent machine-learning based estimators that allow
for the estimation of conditional average treatment effects (CATE) on complex,
high dimensional data. Experimentally, we achieve accurate estimates of
conditional average treatment effects using an ensemble of deep network-based
estimators, including on a challenging simulated Mendelian Randomization
problem.
| stat |
Identification of causal intervention effects under contagion | Defining and identifying causal intervention effects for transmissible
infectious disease outcomes is challenging because a treatment -- such as a
vaccine -- given to one individual may affect the infection outcomes of others.
Epidemiologists have proposed causal estimands to quantify effects of
interventions under contagion using a two-person partnership model. These
simple conceptual models have helped researchers develop causal estimands
relevant to clinical evaluation of vaccine effects. However, many of these
partnership models are formulated under structural assumptions that preclude
realistic infectious disease transmission dynamics, limiting their conceptual
usefulness in defining and identifying causal treatment effects in empirical
intervention trials. In this paper, we propose causal intervention effects in
two-person partnerships under arbitrary infectious disease transmission
dynamics, and give nonparametric identification results showing how effects can
be estimated in empirical trials using time-to-infection or binary outcome
data. The key insight is that contagion is a causal phenomenon that induces
conditional independencies on infection outcomes that can be exploited for the
identification of clinically meaningful causal estimands. These new estimands
are compared to existing quantities, and results are illustrated using a
realistic simulation of an HIV vaccine trial.
| stat |
A direct approach to detection and attribution of climate change | We present here a novel statistical learning approach for detection and
attribution (D&A) of climate change. Traditional optimal D&A studies try to
directly model the observations from model simulations, but practically this is
challenging due to high-dimensionality. Dimension reduction techniques reduce
the dimensionality, typically using empirical orthogonal functions, but as
these techniques are unsupervised, the reduced space considered is somewhat
arbitrary. Here, we propose a supervised approach where we predict a given
external forcing, e.g., anthropogenic forcing, directly from the spatial
pattern of climate variables, and use the predicted forcing as a test statistic
for D&A. We want the prediction to work well even under changes in the
distribution of other external forcings, e.g., solar or volcanic forcings, and
therefore formulate the optimization problem from a distributional robustness
perspective.
| stat |
Adversarial Training for a Continuous Robustness Control Problem in
Power Systems | We propose a new adversarial training approach for injecting robustness when
designing controllers for upcoming cyber-physical power systems. Previous
approaches relying deeply on simulations are not able to cope with the rising
complexity and are too costly when used online in terms of computation budget.
In comparison, our method proves to be computationally efficient online while
displaying useful robustness properties. To do so we model an adversarial
framework, propose the implementation of a fixed opponent policy and test it on
a L2RPN (Learning to Run a Power Network) environment. This environment is a
synthetic but realistic modeling of a cyber-physical system accounting for one
third of the IEEE 118 grid. Using adversarial testing, we analyze the results
of submitted trained agents from the robustness track of the L2RPN competition.
We then further assess the performance of these agents in regards to the
continuous N-1 problem through tailored evaluation metrics. We discover that
some agents trained in an adversarial way demonstrate interesting preventive
behaviors in that regard, which we discuss.
| stat |
Estimation of False Discovery Proportion with Unknown Dependence | Large-scale multiple testing with highly correlated test statistics arises
frequently in many scientific research. Incorporating correlation information
in estimating false discovery proportion has attracted increasing attention in
recent years. When the covariance matrix of test statistics is known, Fan, Han
& Gu (2012) provided a consistent estimate of False Discovery Proportion (FDP)
under arbitrary dependence structure. However, the covariance matrix is often
unknown in many applications and such dependence information has to be
estimated before estimating FDP (Efron, 2010). The estimation accuracy can
greatly affect the convergence result of FDP or even violate its consistency.
In the current paper, we provide methodological modification and theoretical
investigations for estimation of FDP with unknown covariance. First we develop
requirements for estimates of eigenvalues and eigenvectors such that we can
obtain a consistent estimate of FDP. Secondly we give conditions on the
dependence structures such that the estimate of FDP is consistent. Such
dependence structures include sparse covariance matrices, which have been
popularly considered in the contemporary random matrix theory. When data are
sampled from an approximate factor model, which encompasses most practical
situations, we provide a consistent estimate of FDP via exploiting this
specific dependence structure. The results are further demonstrated by
simulation studies and some real data applications.
| stat |
Model Linkage Selection for Cooperative Learning | Rapid developments in data collecting devices and computation platforms
produce an emerging number of learners and data modalities in many scientific
domains. We consider the setting in which each learner holds a pair of
parametric statistical model and a specific data source, with the goal of
integrating information across a set of learners to enhance the prediction
accuracy of a specific learner. One natural way to integrate information is to
build a joint model across a set of learners that shares common parameters of
interest. However, the parameter sharing patterns across a set of learners are
not known a priori. Misspecifying the parameter sharing patterns and the
parametric statistical model for each learner yields a biased estimator and
degrades the prediction accuracy of the joint model. In this paper, we propose
a novel framework for integrating information across a set of learners that is
robust against model misspecification and misspecified parameter sharing
patterns. The main crux is to sequentially incorporates additional learners
that can enhance the prediction accuracy of an existing joint model based on a
user-specified parameter sharing patterns across a set of learners, starting
from a model with one learner. Theoretically, we show that the proposed method
can data-adaptively select the correct parameter sharing patterns based on a
user-specified parameter sharing patterns, and thus enhances the prediction
accuracy of a learner. Extensive numerical studies are performed to evaluate
the performance of the proposed method.
| stat |
Adaptive probabilistic principal component analysis | Using the linear Gaussian latent variable model as a starting point we relax
some of the constraints it imposes by deriving a nonparametric latent feature
Gaussian variable model. This model introduces additional discrete latent
variables to the original structure. The Bayesian nonparametric nature of this
new model allows it to adapt complexity as more data is observed and project
each data point onto a varying number of subspaces. The linear relationship
between the continuous latent and observed variables make the proposed model
straightforward to interpret, resembling a locally adaptive probabilistic PCA
(A-PPCA). We propose two alternative Gibbs sampling procedures for inference in
the new model and demonstrate its applicability on sensor data for passive
health monitoring.
| stat |
Reducing Subspace Models for Large-Scale Covariance Regression | We develop an envelope model for joint mean and covariance regression in the
large $p$, small $n$ setting. In contrast to existing envelope methods, which
improve mean estimates by incorporating estimates of the covariance structure,
we focus on identifying covariance heterogeneity by incorporating information
about mean-level differences. We use a Monte Carlo EM algorithm to identify a
low-dimensional subspace which explains differences in both means and
covariances as a function of covariates, and then use MCMC to estimate the
posterior uncertainty conditional on the inferred low-dimensional subspace. We
demonstrate the utility of our model on a motivating application on the
metabolomics of aging. We also provide R code which can be used to develop and
test other generalizations of the response envelope model.
| stat |
Asymptotic study of stochastic adaptive algorithm in non-convex
landscape | This paper studies some asymptotic properties of adaptive algorithms widely
used in optimization and machine learning, and among them Adagrad and Rmsprop,
which are involved in most of the blackbox deep learning algorithms. Our setup
is the non-convex landscape optimization point of view, we consider a one time
scale parametrization and we consider the situation where these algorithms may
be used or not with mini-batches. We adopt the point of view of stochastic
algorithms and establish the almost sure convergence of these methods when
using a decreasing step-size point of view towards the set of critical points
of the target function. With a mild extra assumption on the noise, we also
obtain the convergence towards the set of minimizer of the function. Along our
study, we also obtain a "convergence rate" of the methods, in the vein of the
works of \cite{GhadimiLan}.
| stat |
Anatomically informed Bayesian spatial priors for fMRI analysis | Existing Bayesian spatial priors for functional magnetic resonance imaging
(fMRI) data correspond to stationary isotropic smoothing filters that may
oversmooth at anatomical boundaries. We propose two anatomically informed
Bayesian spatial models for fMRI data with local smoothing in each voxel based
on a tensor field estimated from a T1-weighted anatomical image. We show that
our anatomically informed Bayesian spatial models results in posterior
probability maps that follow the anatomical structure.
| stat |
Regression modelling of interval censored data based on the adaptive
ridge procedure | A new method for the analysis of time to ankylosis complication on a dataset
of replanted teeth is proposed. In this context of left-censored,
interval-censored and right-censored data, a Cox model with piecewise constant
baseline hazard is introduced. Estimation is carried out with the EM algorithm
by treating the true event times as unobserved variables. This estimation
procedure is shown to produce a block diagonal Hessian matrix of the baseline
parameters. Taking advantage of this interesting feature of the estimation
method a L0 penalised likelihood method is implemented in order to
automatically determine the number and locations of the cuts of the baseline
hazard. This procedure allows to detect specific areas of time where patients
are at greater risks for ankylosis. The method can be directly extended to the
inclusion of exact observations and to a cure fraction. Theoretical results are
obtained which allow to derive statistical inference of the model parameters
from asymptotic likelihood theory. Through simulation studies, the penalisation
technique is shown to provide a good fit of the baseline hazard and precise
estimations of the resulting regression parameters.
| stat |
Spatiotemporal blocking of the bouncy particle sampler for efficient
inference in state space models | We propose a novel blocked version of the continuous-time bouncy particle
sampler of Bouchard- C\^ot\'e et al. [2018] applicable to any differentiable
probability density. Motivated by Singh et al. [2017], we also introduce an
alternative implementation that leads to significant improvement in terms of
effective sample size per second, and furthermore allows for parallelization at
the cost of an extra logarithmic factor. The new algorithms are particularly
efficient for latent state inference in high-dimensional state space models,
where blocking in both space and time is necessary to avoid degeneracy of the
proposal kernel. The efficiency of the blocked bouncy particle sampler, in
comparison with both the standard implementation of the bouncy particle sampler
and the particle Gibbs algorithm of Andrieu et al. [2010], is illustrated
numerically for both simulated data and a challenging real-world financial
dataset.
| stat |
Generalization bounds for graph convolutional neural networks via
Rademacher complexity | This paper aims at studying the sample complexity of graph convolutional
networks (GCNs), by providing tight upper bounds of Rademacher complexity for
GCN models with a single hidden layer. Under regularity conditions, theses
derived complexity bounds explicitly depend on the largest eigenvalue of graph
convolution filter and the degree distribution of the graph. Again, we provide
a lower bound of Rademacher complexity for GCNs to show optimality of our
derived upper bounds. Taking two commonly used examples as representatives, we
discuss the implications of our results in designing graph convolution filters
an graph distribution.
| stat |
A Novel Approach for Semiconductor Etching Process with Inductive Biases | The etching process is one of the most important processes in semiconductor
manufacturing. We have introduced the state-of-the-art deep learning model to
predict the etching profiles. However, the significant problems violating
physics have been found through various techniques such as explainable
artificial intelligence and representation of prediction uncertainty. To
address this problem, this paper presents a novel approach to apply the
inductive biases for etching process. We demonstrate that our approach fits the
measurement faster than physical simulator while following the physical
behavior. Our approach would bring a new opportunity for better etching process
with higher accuracy and lower cost.
| stat |
Pivotal Estimation via Self-Normalization for High-Dimensional Linear
Models with Error in Variables | We propose a new estimator for the high-dimensional linear regression model
with observation error in the design where the number of coefficients is
potentially larger than the sample size. The main novelty of our procedure is
that the choice of penalty parameters is pivotal. The estimator is based on
applying a self-normalization to the constraints that characterize the
estimator. Importantly, we show how to cast the computation of the estimator as
the solution of a convex program with second order cone constraints. This
allows the use of algorithms with theoretical guarantees and reliable
implementation. Under sparsity assumptions, we derive $\ell_q$-rates of
convergence and show that consistency can be achieved even if the number of
regressors exceeds the sample size. We further provide a simple to implement
rule to threshold the estimator that yields a provably sparse estimator with
similar $\ell_2$ and $\ell_1$-rates of convergence. The thresholds are
data-driven and component dependents. Finally, we also study the rates of
convergence of estimators that refit the data based on a selected support with
possible model selection mistakes. In addition to our finite sample theoretical
results that allow for non-i.i.d. data, we also present simulations to compare
the performance of the proposed estimators.
| stat |
High-dimensional GARCH process segmentation with an application to
Value-at-Risk | Models for financial risk often assume that underlying asset returns are
stationary. However, there is strong evidence that multivariate financial time
series entail changes not only in their within-series dependence structure, but
also in the cross-sectional dependence among them. In particular, the stressed
Value-at-Risk of a portfolio, a popularly adopted measure of market risk,
cannot be gauged adequately unless such structural breaks are taken into
account in its estimation. We propose a method for consistent detection of
multiple change points in high-dimensional GARCH panel data set where both
individual GARCH processes and their correlations are allowed to change over
time. We prove its consistency in multiple change point estimation, and
demonstrate its good performance through simulation studies and an application
to the Value-at-Risk problem on a real dataset. Our methodology is implemented
in the R package segMGarch, available from CRAN.
| stat |
A XGBoost risk model via feature selection and Bayesian hyper-parameter
optimization | This paper aims to explore models based on the extreme gradient boosting
(XGBoost) approach for business risk classification. Feature selection (FS)
algorithms and hyper-parameter optimizations are simultaneously considered
during model training. The five most commonly used FS methods including weight
by Gini, weight by Chi-square, hierarchical variable clustering, weight by
correlation, and weight by information are applied to alleviate the effect of
redundant features. Two hyper-parameter optimization approaches, random search
(RS) and Bayesian tree-structured Parzen Estimator (TPE), are applied in
XGBoost. The effect of different FS and hyper-parameter optimization methods on
the model performance are investigated by the Wilcoxon Signed Rank Test. The
performance of XGBoost is compared to the traditionally utilized logistic
regression (LR) model in terms of classification accuracy, area under the curve
(AUC), recall, and F1 score obtained from the 10-fold cross validation. Results
show that hierarchical clustering is the optimal FS method for LR while weight
by Chi-square achieves the best performance in XG-Boost. Both TPE and RS
optimization in XGBoost outperform LR significantly. TPE optimization shows a
superiority over RS since it results in a significantly higher accuracy and a
marginally higher AUC, recall and F1 score. Furthermore, XGBoost with TPE
tuning shows a lower variability than the RS method. Finally, the ranking of
feature importance based on XGBoost enhances the model interpretation.
Therefore, XGBoost with Bayesian TPE hyper-parameter optimization serves as an
operative while powerful approach for business risk modeling.
| stat |
Universal sieve-based strategies for efficient estimation using machine
learning tools | Suppose that we wish to estimate a finite-dimensional summary of one or more
function-valued features of an underlying data-generating mechanism under a
nonparametric model. One approach to estimation is by plugging in flexible
estimates of these features. Unfortunately, in general, such estimators may not
be asymptotically efficient, which often makes these estimators difficult to
use as a basis for inference. Though there are several existing methods to
construct asymptotically efficient plug-in estimators, each such method either
can only be derived using knowledge of efficiency theory or is only valid under
stringent smoothness assumptions. Among existing methods, sieve estimators
stand out as particularly convenient because efficiency theory is not required
in their construction, their tuning parameters can be selected data adaptively,
and they are universal in the sense that the same fits lead to efficient
plug-in estimators for a rich class of estimands. Inspired by these desirable
properties, we propose two novel universal approaches for estimating
function-valued features that can be analyzed using sieve estimation theory.
Compared to traditional sieve estimators, these approaches are valid under more
general conditions on the smoothness of the function-valued features by
utilizing flexible estimates that can be obtained, for example, using machine
learning.
| stat |
Interpretable random forest models through forward variable selection | Random forest is a popular prediction approach for handling high dimensional
covariates. However, it often becomes infeasible to interpret the obtained high
dimensional and non-parametric model. Aiming for obtaining an interpretable
predictive model, we develop a forward variable selection method using the
continuous ranked probability score (CRPS) as the loss function. Our stepwise
procedure leads to a smallest set of variables that optimizes the CRPS risk by
performing at each step a hypothesis test on a significant decrease in CRPS
risk. We provide mathematical motivation for our method by proving that in
population sense the method attains the optimal set. Additionally, we show that
the test is consistent provided that the random forest estimator of a quantile
function is consistent.
In a simulation study, we compare the performance of our method with an
existing variable selection method, for different sample sizes and different
correlation strength of covariates. Our method is observed to have a much lower
false positive rate. We also demonstrate an application of our method to
statistical post-processing of daily maximum temperature forecasts in the
Netherlands. Our method selects about 10% covariates while retaining the same
predictive power.
| stat |
MEMOIR: Multi-class Extreme Classification with Inexact Margin | Multi-class classification with a very large number of classes, or extreme
classification, is a challenging problem from both statistical and
computational perspectives. Most of the classical approaches to multi-class
classification, including one-vs-rest or multi-class support vector machines,
require the exact estimation of the classifier's margin, at both the training
and the prediction steps making them intractable in extreme classification
scenarios. In this paper, we study the impact of computing an approximate
margin using nearest neighbor (ANN) search structures combined with
locality-sensitive hashing (LSH). This approximation allows to dramatically
reduce both the training and the prediction time without a significant loss in
performance. We theoretically prove that this approximation does not lead to a
significant loss of the risk of the model and provide empirical evidence over
five publicly available large scale datasets, showing that the proposed
approach is highly competitive with respect to state-of-the-art approaches on
time, memory and performance measures.
| stat |
Bayesian influence diagnostics using normalizing functional Bregman
divergence | Ideally, any statistical inference should be robust to local influences.
Although there are simple ways to check about leverage points in independent
and linear problems, more complex models require more sophisticated methods.
Kullback-Leiber and Bregman divergences were already applied in Bayesian
inference to measure the isolated impact of each observation in a model. We
extend these ideas to models for dependent data and with non-normal probability
distributions such as time series, spatial models and generalized linear
models. We also propose a strategy to rescale the functional Bregman divergence
to lie in the (0,1) interval thus facilitating interpretation and comparison.
This is accomplished with a minimal computational effort and maintaining all
theoretical properties. For computational efficiency, we take advantage of
Hamiltonian Monte Carlo methods to draw samples from the posterior distribution
of model parameters. The resulting Markov chains are then directly connected
with Bregman calculus, which results in fast computation. We check the
propositions in both simulated and empirical studies.
| stat |
Penalised t-walk MCMC | Handling multimodality that commonly arises from complicated statistical
models remains a challenge. Current Markov chain Monte Carlo (MCMC) methodology
tackling this subject is based on an ensemble of chains targeting a product of
power-tempered distributions. Despite the theoretical validity of such methods,
practical implementations typically suffer from bad mixing and slow convergence
due to the high-computation cost involved. In this work we study novel
extensions of the t-walk algorithm, an existing MCMC method that is inexpensive
and invariant to affine transformations of the state space, for dealing with
multimodal distributions. We acknowledge that the effectiveness of the new
method will be problem dependent and might struggle in complex scenarios; for
such cases we propose a post-processing technique based on pseudo-marginal
theory for combining isolated samples.
| stat |
Quick Learning Mechanism with Cross-Domain Adaptation for Intelligent
Fault Diagnosis | This paper presents a quick learning mechanism for intelligent fault
diagnosis of rotating machines operating under changeable working conditions.
Since real case machines in industries run under different operating
conditions, the deep learning model trained for a laboratory case machine fails
to perform well for the fault diagnosis using recorded data from real case
machines. It poses the need of training a new diagnostic model for the fault
diagnosis of the real case machine under every new working condition.
Therefore, there is a need for a mechanism that can quickly transform the
existing diagnostic model for machines operating under different conditions. we
propose a quick learning method with Net2Net transformation followed by a
fine-tuning method to cancel/minimize the maximum mean discrepancy of the new
data to the previous one. This transformation enables us to create a new
network with any architecture almost ready to be used for the new dataset. The
effectiveness of the proposed fault diagnosis method has been demonstrated on
the CWRU dataset, IMS bearing dataset, and Paderborn university dataset. We
have shown that the diagnostic model trained for CWRU data at zero load can be
used to quickly train another diagnostic model for the CWRU data at different
loads and also for the IMS dataset. Using the dataset provided by Paderborn
university, it has been validated that the diagnostic model trained on
artificially damaged fault dataset can be used for quickly training another
model for real damage dataset.
| stat |
Bayesian joint models for longitudinal and survival data | This paper takes a quick look at Bayesian joint models (BJM) for longitudinal
and survival data. A general formulation for BJM is examined in terms of the
sampling distribution of the longitudinal and survival processes, the
conditional distribution of the random effects and the prior distribution. Next
a basic BJM defined in terms of a mixed linear model and a Cox survival
regression models is discussed and some extensions and other Bayesian topics
are briefly outlined.
| stat |
A Generalized Approximate Control Variate Framework for Multifidelity
Uncertainty Quantification | We describe and analyze a variance reduction approach for Monte Carlo (MC)
sampling that accelerates the estimation of statistics of computationally
expensive simulation models using an ensemble of models with lower cost. These
lower cost models --- which are typically lower fidelity with unknown
statistics --- are used to reduce the variance in statistical estimators
relative to a MC estimator with equivalent cost. We derive the conditions under
which our proposed approximate control variate framework recovers existing
multi-model variance reduction schemes as special cases. We demonstrate that
these existing strategies use recursive sampling strategies, and as a result,
their maximum possible variance reduction is limited to that of a control
variate algorithm that uses only a single low-fidelity model with known mean.
This theoretical result holds regardless of the number of low-fidelity models
and/or samples used to build the estimator. We then derive new sampling
strategies within our framework that circumvent this limitation to make
efficient use of all available information sources. In particular, we
demonstrate that a significant gap can exist, of orders of magnitude in some
cases, between the variance reduction achievable by using a single low-fidelity
model and our non-recursive approach. We also present initial sample allocation
approaches for exploiting this gap. They yield the greatest benefit when
augmenting the high-fidelity model evaluations is impractical because, for
instance, they arise from a legacy database. Several analytic examples and an
example with a hyperbolic PDE describing elastic wave propagation in
heterogeneous media are used to illustrate the main features of the
methodology.
| stat |
Bayesian Allocation Model: Inference by Sequential Monte Carlo for
Nonnegative Tensor Factorizations and Topic Models using Polya Urns | We introduce a dynamic generative model, Bayesian allocation model (BAM),
which establishes explicit connections between nonnegative tensor factorization
(NTF), graphical models of discrete probability distributions and their
Bayesian extensions, and the topic models such as the latent Dirichlet
allocation. BAM is based on a Poisson process, whose events are marked by using
a Bayesian network, where the conditional probability tables of this network
are then integrated out analytically. We show that the resulting marginal
process turns out to be a Polya urn, an integer valued self-reinforcing
process. This urn processes, which we name a Polya-Bayes process, obey certain
conditional independence properties that provide further insight about the
nature of NTF. These insights also let us develop space efficient simulation
algorithms that respect the potential sparsity of data: we propose a class of
sequential importance sampling algorithms for computing NTF and approximating
their marginal likelihood, which would be useful for model selection. The
resulting methods can also be viewed as a model scoring method for topic models
and discrete Bayesian networks with hidden variables. The new algorithms have
favourable properties in the sparse data regime when contrasted with
variational algorithms that become more accurate when the total sum of the
elements of the observed tensor goes to infinity. We illustrate the performance
on several examples and numerically study the behaviour of the algorithms for
various data regimes.
| stat |
On the Difficulty of Unbiased Alpha Divergence Minimization | Several approximate inference algorithms have been proposed to minimize an
alpha-divergence between an approximating distribution and a target
distribution. Many of these algorithms introduce bias, the magnitude of which
is poorly understood. Other algorithms are unbiased. These often seem to suffer
from high variance, but again, little is rigorously known. In this work we
study unbiased methods for alpha-divergence minimization through the
Signal-to-Noise Ratio (SNR) of the gradient estimator. We study several
representative scenarios where strong analytical results are possible, such as
fully-factorized or Gaussian distributions. We find that when alpha is not
zero, the SNR worsens exponentially in the dimensionality of the problem. This
casts doubt on the practicality of these methods. We empirically confirm these
theoretical results.
| stat |
Robust Similarity and Distance Learning via Decision Forests | Canonical distances such as Euclidean distance often fail to capture the
appropriate relationships between items, subsequently leading to subpar
inference and prediction. Many algorithms have been proposed for automated
learning of suitable distances, most of which employ linear methods to learn a
global metric over the feature space. While such methods offer nice theoretical
properties, interpretability, and computationally efficient means for
implementing them, they are limited in expressive capacity. Methods which have
been designed to improve expressiveness sacrifice one or more of the nice
properties of the linear methods. To bridge this gap, we propose a highly
expressive novel decision forest algorithm for the task of distance learning,
which we call Similarity and Metric Random Forests (SMERF). We show that the
tree construction procedure in SMERF is a proper generalization of standard
classification and regression trees. Thus, the mathematical driving forces of
SMERF are examined via its direct connection to regression forests, for which
theory has been developed. Its ability to approximate arbitrary distances and
identify important features is empirically demonstrated on simulated data sets.
Last, we demonstrate that it accurately predicts links in networks.
| stat |
Visualizing Outliers in High Dimensional Functional Data for Task fMRI
data exploration | Task-based functional magnetic resonance imaging (task fMRI) is a
non-invasive technique that allows identifying brain regions whose activity
changes when individuals are asked to perform a given task. This contributes to
the understanding of how the human brain is organized in functionally distinct
subdivisions. Task fMRI experiments from high-resolution scans provide hundred
of thousands of longitudinal signals for each individual, corresponding to
measurements of brain activity over each voxel of the brain along the duration
of the experiment. In this context, we propose some visualization techniques
for high dimensional functional data relying on depth-based notions that allow
for computationally efficient 2-dim representations of tfMRI data and that shed
light on sample composition, outlier presence and individual variability. We
believe that this step is crucial previously to any inferential approach
willing to identify neuroscientific patterns across individuals, tasks and
brain regions. We illustrate the proposed technique through a simulation study
and demonstrate its application on a motor and language task fMRI experiment.
| stat |
A Bias-Corrected Estimator for the Crosswise Model with Inattentive
Respondents | The crosswise model is an increasingly popular survey technique to elicit
candid answers from respondents on sensitive questions. Recent studies,
however, point out that in the presence of inattentive respondents, the
conventional estimator of the prevalence of a sensitive attribute is biased
toward 0.5. To remedy this problem, we propose a simple design-based bias
correction using an anchor question that has a sensitive item with known
prevalence. We demonstrate that we can easily estimate and correct for the bias
arising from inattentive respondents without measuring individual-level
attentiveness. We also offer several useful extensions of our estimator,
including a sensitivity analysis for the conventional estimator, a strategy for
weighting, a framework for multivariate regressions in which a latent sensitive
trait is used as an outcome or a predictor, and tools for power analysis and
parameter selection. Our method can be easily implemented through our
open-source software, cWise.
| stat |
Heterogeneous large datasets integration using Bayesian factor
regression | Two key challenges in modern statistical applications are the large amount of
information recorded per individual, and that such data are often not collected
all at once but in batches. These batch effects can be complex, causing
distortions in both mean and variance. We propose a novel sparse latent factor
regression model to integrate such heterogeneous data. The model provides a
tool for data exploration via dimensionality reduction while correcting for a
range of batch effects. We study the use of several sparse priors (local and
non-local) to learn the dimension of the latent factors. Our model is fitted in
a deterministic fashion by means of an EM algorithm for which we derive
closed-form updates, contributing a novel scalable algorithm for non-local
priors of interest beyond the immediate scope of this paper. We present several
examples, with a focus on bioinformatics applications. Our results show an
increase in the accuracy of the dimensionality reduction, with non-local priors
substantially improving the reconstruction of factor cardinality, as well as
the need to account for batch effects to obtain reliable results. Our model
provides a novel approach to latent factor regression that balances sparsity
with sensitivity and is highly computationally efficient.
| stat |
Correlated pseudo-marginal schemes for time-discretised stochastic
kinetic models | The challenging problem of conducting fully Bayesian inference for the
reaction rate constants governing stochastic kinetic models (SKMs) is
considered. Given the challenges underlying this problem, the Markov jump
process representation is routinely replaced by an approximation based on a
suitable time discretisation of the system of interest. Improving the accuracy
of these schemes amounts to using an ever finer discretisation level, which in
the context of the inference problem, requires integrating over the uncertainty
in the process at a predetermined number of intermediate times between
observations. Pseudo-marginal Metropolis-Hastings schemes are increasingly
used, since for a given discretisation level, the observed data likelihood can
be unbiasedly estimated using a particle filter. When observations are
particularly informative an auxiliary particle filter can be implemented, by
employing an appropriate construct to push the state particles towards the
observations in a sensible way. Recent work in state-space settings has shown
how the pseudo-marginal approach can be made much more efficient by correlating
the underlying pseudo-random numbers used to form the likelihood estimate at
the current and proposed values of the unknown parameters. We extend this
approach to the time-discretised SKM framework by correlating the innovations
that drive the auxiliary particle filter. We find that the resulting approach
offers substantial gains in efficiency over a standard implementation.
| stat |
A Single SMC Sampler on MPI that Outperforms a Single MCMC Sampler | Markov Chain Monte Carlo (MCMC) is a well-established family of algorithms
which are primarily used in Bayesian statistics to sample from a target
distribution when direct sampling is challenging. Single instances of MCMC
methods are widely considered hard to parallelise in a problem-agnostic fashion
and hence, unsuitable to meet both constraints of high accuracy and high
throughput. Sequential Monte Carlo (SMC) Samplers can address the same problem,
but are parallelisable: they share with Particle Filters the same key tasks and
bottleneck. Although a rich literature already exists on MCMC methods, SMC
Samplers are relatively underexplored, such that no parallel implementation is
currently available. In this paper, we first propose a parallel MPI version of
the SMC Sampler, including an optimised implementation of the bottleneck, and
then compare it with single-core Metropolis-Hastings. The goal is to show that
SMC Samplers may be a promising alternative to MCMC methods with high potential
for future improvements. We demonstrate that a basic SMC Sampler with 512 cores
is up to 85 times faster or up to 8 times more accurate than
Metropolis-Hastings.
| stat |
A comparative study of scoring systems by simulations | Scoring rules aggregate individual rankings by assigning some points to each
position in each ranking such that the total sum of points provides the overall
ranking of the alternatives. They are widely used in sports competitions
consisting of multiple contests. We study the tradeoff between two risks in
this setting: (1) the threat of early clinch when the title has been clinched
before the last contest(s) of the competition take place; (2) the danger of
winning the competition without finishing first in any contest. In particular,
four historical points scoring systems of the Formula One World Championship
are compared with the family of geometric scoring rules that have favourable
axiomatic properties. The formers are found to be competitive or even better.
The current scheme seems to be a reasonable compromise in optimising the above
goals. Our results shed more light on the evolution of the Formula One points
scoring systems and contribute to the issue of choosing the set of point
values.
| stat |
CMS Sematrix: A Tool to Aid the Development of Clinical Quality Measures
(CQMs) | As part of the effort to improve quality and to reduce national healthcare
costs, the Centers for Medicare and Medicaid Services (CMS) are responsible for
creating and maintaining an array of clinical quality measures (CQMs) for
assessing healthcare structure, process, outcome, and patient experience across
various conditions, clinical specialties, and settings. The development and
maintenance of CQMs involves substantial and ongoing evaluation of the evidence
on the measure's properties: importance, reliability, validity, feasibility,
and usability. As such, CMS conducts monthly environmental scans of the
published clinical and health service literature. Conducting time consuming,
exhaustive evaluations of the ever-changing healthcare literature presents one
of the largest challenges to an evidence-based approach to healthcare quality
improvement. Thus, it is imperative to leverage automated techniques to aid CMS
in the identification of clinical and health services literature relevant to
CQMs. Additionally, the estimated labor hours and related cost savings of using
CMS Sematrix compared to a traditional literature review are roughly 818 hours
and 122,000 dollars for a single monthly environmental scan.
| stat |
An Energy Approach to the Solution of Partial Differential Equations in
Computational Mechanics via Machine Learning: Concepts, Implementation and
Applications | Partial Differential Equations (PDE) are fundamental to model different
phenomena in science and engineering mathematically. Solving them is a crucial
step towards a precise knowledge of the behaviour of natural and engineered
systems. In general, in order to solve PDEs that represent real systems to an
acceptable degree, analytical methods are usually not enough. One has to resort
to discretization methods. For engineering problems, probably the best known
option is the finite element method (FEM). However, powerful alternatives such
as mesh-free methods and Isogeometric Analysis (IGA) are also available. The
fundamental idea is to approximate the solution of the PDE by means of
functions specifically built to have some desirable properties. In this
contribution, we explore Deep Neural Networks (DNNs) as an option for
approximation. They have shown impressive results in areas such as visual
recognition. DNNs are regarded here as function approximation machines. There
is great flexibility to define their structure and important advances in the
architecture and the efficiency of the algorithms to implement them make DNNs a
very interesting alternative to approximate the solution of a PDE. We
concentrate in applications that have an interest for Computational Mechanics.
Most contributions that have decided to explore this possibility have adopted a
collocation strategy. In this contribution, we concentrate in mechanical
problems and analyze the energetic format of the PDE. The energy of a
mechanical system seems to be the natural loss function for a machine learning
method to approach a mechanical problem. As proofs of concept, we deal with
several problems and explore the capabilities of the method for applications in
engineering.
| stat |
A Unifying Bayesian View of Continual Learning | Some machine learning applications require continual learning - where data
comes in a sequence of datasets, each is used for training and then permanently
discarded. From a Bayesian perspective, continual learning seems
straightforward: Given the model posterior one would simply use this as the
prior for the next task. However, exact posterior evaluation is intractable
with many models, especially with Bayesian neural networks (BNNs). Instead,
posterior approximations are often sought. Unfortunately, when posterior
approximations are used, prior-focused approaches do not succeed in evaluations
designed to capture properties of realistic continual learning use cases. As an
alternative to prior-focused methods, we introduce a new approximate Bayesian
derivation of the continual learning loss. Our loss does not rely on the
posterior from earlier tasks, and instead adapts the model itself by changing
the likelihood term. We call these approaches likelihood-focused. We then
combine prior- and likelihood-focused methods into one objective, tying the two
views together under a single unifying framework of approximate Bayesian
continual learning.
| stat |
Sequential Aggregation of Probabilistic Forecasts -- Applicaton to Wind
Speed Ensemble Forecasts | In the field of numerical weather prediction (NWP), the probabilistic
distribution of the future state of the atmosphere is sampled with
Monte-Carlo-like simulations, called ensembles. These ensembles have
deficiencies (such as conditional biases) that can be corrected thanks to
statistical post-processing methods. Several ensembles exist and may be
corrected with different statistiscal methods. A further step is to combine
these raw or post-processed ensembles. The theory of prediction with expert
advice allows us to build combination algorithms with theoretical guarantees on
the forecast performance. This article adapts this theory to the case of
probabilistic forecasts issued as step-wise cumulative distribution functions
(CDF). The theory is applied to wind speed forecasting, by combining several
raw or post-processed ensembles, considered as CDFs. The second goal of this
study is to explore the use of two forecast performance criteria: the Continous
ranked probability score (CRPS) and the Jolliffe-Primo test. Comparing the
results obtained with both criteria leads to reconsidering the usual way to
build skillful probabilistic forecasts, based on the minimization of the CRPS.
Minimizing the CRPS does not necessarily produce reliable forecasts according
to the Jolliffe-Primo test. The Jolliffe-Primo test generally selects reliable
forecasts, but could lead to issuing suboptimal forecasts in terms of CRPS. It
is proposed to use both criterion to achieve reliable and skillful
probabilistic forecasts.
| stat |
A multinomial quadrivariate D-vine copula mixed model for meta-analysis
of diagnostic studies in the presence of non-evaluable subjects | Diagnostic test accuracy studies observe the result of a gold standard
procedure that defines the presence or absence of a disease and the result of a
diagnostic test. They typically report the number of true positives, false
positives, true negatives and false negatives. However, diagnostic test
outcomes can also be either non-evaluable positives or non-evaluable negatives.
We propose a novel model for meta-analysis of diagnostic studies in the
presence of non-evaluable outcomes that assumes independent multinomial
distributions for the true and non-evaluable positives, and, the true and non
evaluable negatives, conditional on the latent sensitivity, specificity,
probability of non-evaluable positives and probability of non-evaluable
negatives in each study. For the random effects distribution of the latent
proportions, we employ a drawable vine copula that can successively model the
dependence in the joint tails. Our methodology is demonstrated with an
extensive simulation study and applied to data from diagnostic accuracy studies
of coronary computed tomography angiography for the detection of coronary
artery disease. The comparison of our method with the existing approaches
yields findings in the real data application that change the current
conclusions.
| 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.