title
stringlengths 6
244
| abstract
stringlengths 19
5.09k
| label
stringclasses 10
values |
|---|---|---|
On In-network learning. A Comparative Study with Federated and Split
Learning | In this paper, we consider a problem in which distributively extracted
features are used for performing inference in wireless networks. We elaborate
on our proposed architecture, which we herein refer to as "in-network
learning", provide a suitable loss function and discuss its optimization using
neural networks. We compare its performance with both Federated- and Split
learning; and show that this architecture offers both better accuracy and
bandwidth savings.
| stat |
Sparse and Smooth Signal Estimation: Convexification of L0 Formulations | Signal estimation problems with smoothness and sparsity priors can be
naturally modeled as quadratic optimization with $\ell_0$-"norm" constraints.
Since such problems are non-convex and hard-to-solve, the standard approach is,
instead, to tackle their convex surrogates based on $\ell_1$-norm relaxations.
In this paper, we propose a new iterative (convex) conic quadratic relaxations
that exploit not only the $\ell_0$-"norm" terms, but also the fitness and
smoothness functions. The iterative convexification approach substantially
closes the gap between the $\ell_0$-"norm" and its $\ell_1$ surrogate. These
stronger relaxations lead to significantly better estimators than $\ell_1$-norm
approaches and also allow one to utilize affine sparsity priors. In addition,
the parameters of the model and the resulting estimators are easily
interpretable. Experiments with a tailored Lagrangian decomposition method
indicate that the proposed iterative convex relaxations \rev{yield solutions
within 1\% of the exact $\ell_0$ approach, and can tackle instances with up to
100,000 variables under one minute.
| stat |
Bayesian time-aligned factor analysis of paired multivariate time series | Many modern data sets require inference methods that can estimate the shared
and individual-specific components of variability in collections of matrices
that change over time. Promising methods have been developed to analyze these
types of data in static cases, but very few approaches are available for
dynamic settings. To address this gap, we consider novel models and inference
methods for pairs of matrices in which the columns correspond to multivariate
observations at different time points. In order to characterize common and
individual features, we propose a Bayesian dynamic factor modeling framework
called Time Aligned Common and Individual Factor Analysis (TACIFA) that
includes uncertainty in time alignment through an unknown warping function. We
provide theoretical support for the proposed model, showing identifiability and
posterior concentration. The structure enables efficient computation through a
Hamiltonian Monte Carlo (HMC) algorithm. We show excellent performance in
simulations, and illustrate the method through application to a social
synchrony experiment.
| stat |
Peer groups for organisational learning: clustering with practical
constraints | Peer-grouping is used in many sectors for organisational learning, policy
implementation, and benchmarking. Clustering provides a statistical,
data-driven method for constructing meaningful peer groups, but peer groups
must be compatible with business constraints such as size and stability
considerations. Additionally, statistical peer groups are constructed from many
different variables, and can be difficult to understand, especially for
non-statistical audiences. We developed methodology to apply business
constraints to clustering solutions and allow the decision-maker to choose the
balance between statistical goodness-of-fit and conformity to business
constraints. Several tools were utilised to identify complex distinguishing
features in peer groups, and a number of visualisations are developed to
explain high-dimensional clusters for non-statistical audiences. In a case
study where peer group size was required to be small ($\leq 100$ members), we
applied constrained clustering to a noisy high-dimensional data-set over two
subsequent years, ensuring that the clusters were sufficiently stable between
years. Our approach not only satisfied clustering constraints on the test data,
but maintained an almost monotonic negative relationship between
goodness-of-fit and stability between subsequent years. We demonstrated in the
context of the case study how distinguishing features between clusters can be
communicated clearly to different stakeholders with substantial and limited
statistical knowledge.
| stat |
Introducing an Explicit Symplectic Integration Scheme for Riemannian
Manifold Hamiltonian Monte Carlo | We introduce a recent symplectic integration scheme derived for solving
physically motivated systems with non-separable Hamiltonians. We show its
relevance to Riemannian manifold Hamiltonian Monte Carlo (RMHMC) and provide an
alternative to the currently used generalised leapfrog symplectic integrator,
which relies on solving multiple fixed point iterations to convergence. Via
this approach, we are able to reduce the number of higher-order derivative
calculations per leapfrog step. We explore the implications of this integrator
and demonstrate its efficacy in reducing the computational burden of RMHMC. Our
code is provided in a new open-source Python package, hamiltorch.
| stat |
f-SAEM: A fast Stochastic Approximation of the EM algorithm for
nonlinear mixed effects models | The ability to generate samples of the random effects from their conditional
distributions is fundamental for inference in mixed effects models. Random walk
Metropolis is widely used to perform such sampling, but this method is known to
converge slowly for medium dimensional problems, or when the joint structure of
the distributions to sample is spatially heterogeneous. The main contribution
consists of an independent Metropolis-Hastings (MH) algorithm based on a
multidimensional Gaussian proposal that takes into account the joint
conditional distribution of the random effects and does not require any tuning.
Indeed, this distribution is automatically obtained thanks to a Laplace
approximation of the incomplete data model. Such approximation is shown to be
equivalent to linearizing the structural model in the case of continuous data.
Numerical experiments based on simulated and real data illustrate the
performance of the proposed methods. For fitting nonlinear mixed effects
models, the suggested MH algorithm is efficiently combined with a stochastic
approximation version of the EM algorithm for maximum likelihood estimation of
the global parameters.
| stat |
Estimating Bayes factors from minimal ANOVA summaries for
repeated-measures designs | In this paper, I develop a formula for estimating Bayes factors directly from
minimal summary statistics produced in repeated measures analysis of variance
designs. The formula, which requires knowing only the $F$-statistic, the number
of subjects, and the number of repeated measurements per subject, is based on
the BIC approximation of the Bayes factor, a common default method for Bayesian
computation with linear models. In addition to providing computational
examples, I report a simulation study in which I demonstrate that the formula
compares favorably to a recently developed, more complex method that accounts
for correlation between repeated measurements. The minimal BIC method provides
a simple way for researchers to estimate Bayes factors from a minimal set of
summary statistics, giving users a powerful index for estimating the evidential
value of not only their own data, but also the data reported in published
studies.
| stat |
Bayesian sparse convex clustering via global-local shrinkage priors | Sparse convex clustering is to cluster observations and conduct variable
selection simultaneously in the framework of convex clustering. Although a
weighted $L_1$ norm is usually employed for the regularization term in sparse
convex clustering, its use increases the dependence on the data and reduces the
estimation accuracy if the sample size is not sufficient. To tackle these
problems, this paper proposes a Bayesian sparse convex clustering method based
on the ideas of Bayesian lasso and global-local shrinkage priors. We introduce
Gibbs sampling algorithms for our method using scale mixtures of normal
distributions. The effectiveness of the proposed methods is shown in simulation
studies and a real data analysis.
| stat |
An Unified Semiparametric Approach to Model Lifetime Data with Crossing
Survival Curves | The proportional hazards (PH), proportional odds (PO) and accelerated failure
time (AFT) models have been widely used in different applications of survival
analysis. Despite their popularity, these models are not suitable to handle
lifetime data with crossing survival curves. In 2005, Yang and Prentice
proposed a semiparametric two-sample strategy (YP model), including the PH and
PO frameworks as particular cases, to deal with this type of data. Assuming a
general regression setting, the present paper proposes an unified approach to
fit the YP model by employing Bernstein polynomials to manage the baseline
hazard and odds under both the frequentist and Bayesian frameworks. The use of
the Bernstein polynomials has some advantages: it allows for uniform
approximation of the baseline distribution, it leads to closed-form expressions
for all baseline functions, it simplifies the inference procedure, and the
presence of a continuous survival function allows a more accurate estimation of
the crossing survival time. Extensive simulation studies are carried out to
evaluate the behavior of the models. The analysis of a clinical trial data set,
related to non-small-cell lung cancer, is also developed as an illustration.
Our findings indicate that assuming the usual PH model, ignoring the existing
crossing survival feature in the real data, is a serious mistake with
implications for those patients in the initial stage of treatment.
| stat |
Graphical models for circular variables | Graphical models are a key class of probabilistic models for studying the
conditional independence structure of a set of random variables. Circular
variables are special variables, characterized by periodicity, arising in
several contexts and fields. However, models for studying the
dependence/independence structure of circular variables are under-explored.
This paper analyses three multivariate circular distributions, the von Mises,
the Wrapped Normal and the Inverse Stereographic distributions, focusing on
their properties concerning conditional independence. For each one of these
distributions, we discuss the main properties related to conditional
independence and introduce suitable classes of graphical models. The usefulness
of the proposed models is shown by modelling the conditional independence among
dihedral angles characterizing the three-dimensional structure of some
proteins.
| stat |
Difference-in-Differences for Ordinal Outcomes: Application to the
Effect of Mass Shootings on Attitudes toward Gun Control | The difference-in-differences (DID) design is widely used in observational
studies to estimate the causal effect of a treatment when repeated observations
over time are available. Yet, almost all existing methods assume linearity in
the potential outcome (parallel trends assumption) and target the additive
effect. In social science research, however, many outcomes of interest are
measured on an ordinal scale. This makes the linearity assumption inappropriate
because the difference between two ordinal potential outcomes is not well
defined. In this paper, I propose a method to draw causal inferences for
ordinal outcomes under the DID design. Unlike existing methods, the proposed
method utilizes the latent variable framework to handle the non-numeric nature
of the outcome, enabling identification and estimation of causal effects based
on the assumption on the quantile of the latent continuous variable. The paper
also proposes an equivalence-based test to assess the plausibility of the key
identification assumption when additional pre-treatment periods are available.
The proposed method is applied to a study estimating the causal effect of mass
shootings on the public's support for gun control. I find little evidence for a
uniform shift toward pro-gun control policies as found in the previous study,
but find that the effect is concentrated on left-leaning respondents who
experienced the shooting for the first time in more than a decade.
| stat |
Random Machines: A bagged-weighted support vector model with free kernel
choice | Improvement of statistical learning models in order to increase efficiency in
solving classification or regression problems is still a goal pursued by the
scientific community. In this way, the support vector machine model is one of
the most successful and powerful algorithms for those tasks. However, its
performance depends directly from the choice of the kernel function and their
hyperparameters. The traditional choice of them, actually, can be
computationally expensive to do the kernel choice and the tuning processes. In
this article, it is proposed a novel framework to deal with the kernel function
selection called Random Machines. The results improved accuracy and reduced
computational time. The data study was performed in simulated data and over 27
real benchmarking datasets.
| stat |
Stochastic Coordinate Minimization with Progressive Precision for
Stochastic Convex Optimization | A framework based on iterative coordinate minimization (CM) is developed for
stochastic convex optimization. Given that exact coordinate minimization is
impossible due to the unknown stochastic nature of the objective function, the
crux of the proposed optimization algorithm is an optimal control of the
minimization precision in each iteration. We establish the optimal precision
control and the resulting order-optimal regret performance for strongly convex
and separably nonsmooth functions. An interesting finding is that the optimal
progression of precision across iterations is independent of the
low-dimensional CM routine employed, suggesting a general framework for
extending low-dimensional optimization routines to high-dimensional problems.
The proposed algorithm is amenable to online implementation and inherits the
scalability and parallelizability properties of CM for large-scale
optimization. Requiring only a sublinear order of message exchanges, it also
lends itself well to distributed computing as compared with the alternative
approach of coordinate gradient descent.
| stat |
Orthogonal Series Density Estimation for Complex Surveys | We propose an orthogonal series density estimator for complex surveys, where
samples are neither independent nor identically distributed. The proposed
estimator is proved to be design-unbiased and asymptotically design-consistent.
The asymptotic normality is proved under both design and combined spaces. Two
data driven estimators are proposed based on the proposed oracle estimator. We
show the efficiency of the proposed estimators in simulation studies. A real
survey data example is provided for an illustration.
| stat |
APIK: Active Physics-Informed Kriging Model with Partial Differential
Equations | Kriging (or Gaussian process regression) is a popular machine learning method
for its flexibility and closed-form prediction expressions. However, one of the
key challenges in applying kriging to engineering systems is that the available
measurement data is scarce due to the measurement limitations and high sensing
costs. On the other hand, physical knowledge of the engineering system is often
available and represented in the form of partial differential equations (PDEs).
We present in this work a PDE Informed Kriging model (PIK), which introduces
PDE information via a set of PDE points and conducts posterior prediction
similar to the standard kriging method. The proposed PIK model can incorporate
physical knowledge from both linear and nonlinear PDEs. To further improve
learning performance, we propose an Active PIK framework (APIK) that designs
PDE points to leverage the PDE information based on the PIK model and
measurement data. The selected PDE points not only explore the whole input
space but also exploit the locations where the PDE information is critical in
reducing predictive uncertainty. Finally, an expectation-maximization algorithm
is developed for parameter estimation. We demonstrate the effectiveness of APIK
in two synthetic examples, a shock wave case study, and a laser heating case
study.
| stat |
Beyond the Chinese Restaurant and Pitman-Yor processes: Statistical
Models with Double Power-law Behavior | Bayesian nonparametric approaches, in particular the Pitman-Yor process and
the associated two-parameter Chinese Restaurant process, have been successfully
used in applications where the data exhibit a power-law behavior. Examples
include natural language processing, natural images or networks. There is also
growing empirical evidence that some datasets exhibit a two-regime power-law
behavior: one regime for small frequencies, and a second regime, with a
different exponent, for high frequencies. In this paper, we introduce a class
of completely random measures which are doubly regularly-varying. Contrary to
the Pitman-Yor process, we show that when completely random measures in this
class are normalized to obtain random probability measures and associated
random partitions, such partitions exhibit a double power-law behavior. We
discuss in particular three models within this class: the beta prime process
(Broderick et al. (2015, 2018), a novel process called generalized BFRY
process, and a mixture construction. We derive efficient Markov chain Monte
Carlo algorithms to estimate the parameters of these models. Finally, we show
that the proposed models provide a better fit than the Pitman-Yor process on
various datasets.
| stat |
spsurv: An R package for semi-parametric survival analysis | Software development innovations and advances in computing have enabled more
complex and less costly computations in medical research (survival analysis),
engineering studies (reliability analysis), and social sciences event analysis
(historical analysis). As a result, many semi-parametric modeling efforts
emerged when it comes to time-to-event data analysis. In this context, this
work presents a flexible Bernstein polynomial (BP) based framework for survival
data modeling. This innovative approach is applied to existing families of
models such as proportional hazards (PH), proportional odds (PO), and
accelerated failure time (AFT) models to estimate unknown baseline functions.
Along with this contribution, this work also presents new automated routines in
R, taking advantage of algorithms available in Stan. The proposed computation
routines are tested and explored through simulation studies based on artificial
datasets. The tools implemented to fit the proposed statistical models are
combined and organized in an R package. Also, the BP based proportional hazards
(BPPH), proportional odds (BPPO), and accelerated failure time (BPAFT) models
are illustrated in real applications related to cancer trial data using maximum
likelihood (ML) estimation and Markov chain Monte Carlo (MCMC) methods.
| stat |
Multiway clustering via tensor block models | We consider the problem of identifying multiway block structure from a large
noisy tensor. Such problems arise frequently in applications such as genomics,
recommendation system, topic modeling, and sensor network localization. We
propose a tensor block model, develop a unified least-square estimation, and
obtain the theoretical accuracy guarantees for multiway clustering. The
statistical convergence of the estimator is established, and we show that the
associated clustering procedure achieves partition consistency. A sparse
regularization is further developed for identifying important blocks with
elevated means. The proposal handles a broad range of data types, including
binary, continuous, and hybrid observations. Through simulation and application
to two real datasets, we demonstrate the outperformance of our approach over
previous methods.
| stat |
Autoregressive Hidden Markov Models with partial knowledge on latent
space applied to aero-engines prognostics | [This paper was initially published in PHME conference in 2016, selected for
further publication in International Journal of Prognostics and Health
Management.]
This paper describes an Autoregressive Partially-hidden Markov model (ARPHMM)
for fault detection and prognostics of equipments based on sensors' data. It is
a particular dynamic Bayesian network that allows to represent the dynamics of
a system by means of a Hidden Markov Model (HMM) and an autoregressive (AR)
process. The Markov chain assumes that the system is switching back and forth
between internal states while the AR process ensures a temporal coherence on
sensor measurements. A sound learning procedure of standard ARHMM based on
maximum likelihood allows to iteratively estimate all parameters
simultaneously. This paper suggests a modification of the learning procedure
considering that one may have prior knowledge about the structure which becomes
partially hidden. The integration of the prior is based on the Theory of
Weighted Distributions which is compatible with the Expectation-Maximization
algorithm in the sense that the convergence properties are still satisfied. We
show how to apply this model to estimate the remaining useful life based on
health indicators. The autoregressive parameters can indeed be used for
prediction while the latent structure can be used to get information about the
degradation level. The interest of the proposed method for prognostics and
health assessment is demonstrated on CMAPSS datasets.
| stat |
Measuring Causality: The Science of Cause and Effect | Determining and measuring cause-effect relationships is fundamental to most
scientific studies of natural phenomena. The notion of causation is distinctly
different from correlation which only looks at association of trends or
patterns in measurements. In this article, we review different notions of
causality and focus especially on measuring causality from time series data.
Causality testing finds numerous applications in diverse disciplines such as
neuroscience, econometrics, climatology, physics and artificial intelligence.
| stat |
Demand forecasting in supply chain: The impact of demand volatility in
the presence of promotion | The demand for a particular product or service is typically associated with
different uncertainties that can make them volatile and challenging to predict.
Demand unpredictability is one of the managers' concerns in the supply chain
that can cause large forecasting errors, issues in the upstream supply chain
and impose unnecessary costs. We investigate 843 real demand time series with
different values of coefficient of variations (CoV) where promotion causes
volatility over the entire demand series. In such a case, forecasting demand
for different CoV require different models to capture the underlying behavior
of demand series and pose significant challenges due to very different and
diverse demand behavior. We decompose demand into baseline and promotional
demand and propose a hybrid model to forecast demand. Our results indicate that
our proposed hybrid model generates robust and accurate forecast across series
with different levels of volatilities. We stress the necessity of decomposition
for volatile demand series. We also model demand series with a number of well
known statistical and machine learning (ML) models to investigate their
performance empirically. We found that ARIMA with covariate (ARIMAX) works well
to forecast volatile demand series, but exponential smoothing with covariate
(ETSX) has a poor performance. Support vector regression (SVR) and dynamic
linear regression (DLR) models generate robust forecasts across different
categories of demands with different CoV values.
| stat |
A Formal Causal Interpretation of the Case-Crossover Design | The case-crossover design (Maclure, 1991) is widely used in epidemiology and
other fields to study causal effects of transient treatments on acute outcomes.
However, its validity and causal interpretation have only been justified under
informal conditions. Here, we place the design in a formal counterfactual
framework for the first time. Doing so helps to clarify its assumptions and
interpretation. In particular, when the treatment effect is non-null, we
identify a previously unnoticed bias arising from common causes of the outcome
at different person-times. We analytically characterize the direction and size
of this bias and demonstrate its potential importance with a simulation. We
also use our derivation of the limit of the case-crossover estimator to analyze
its sensitivity to treatment effect heterogeneity, a violation of one of the
informal criteria for validity. The upshot of this work for practitioners is
that, while the case-crossover design can be useful for testing the causal null
hypothesis in the presence of baseline confounders, extra caution is warranted
when using the case-crossover design for point estimation of causal effects.
| stat |
Adversarial Example Decomposition | Research has shown that widely used deep neural networks are vulnerable to
carefully crafted adversarial perturbations. Moreover, these adversarial
perturbations often transfer across models. We hypothesize that adversarial
weakness is composed of three sources of bias: architecture, dataset, and
random initialization. We show that one can decompose adversarial examples into
an architecture-dependent component, data-dependent component, and
noise-dependent component and that these components behave intuitively. For
example, noise-dependent components transfer poorly to all other models, while
architecture-dependent components transfer better to retrained models with the
same architecture. In addition, we demonstrate that these components can be
recombined to improve transferability without sacrificing efficacy on the
original model.
| stat |
Towards a Better Understanding and Regularization of GAN Training
Dynamics | Generative adversarial networks (GANs) are notoriously difficult to train and
the reasons underlying their (non-)convergence behaviors are still not
completely understood. By first considering a simple yet representative GAN
example, we mathematically analyze its local convergence behavior in a
non-asymptotic way. Furthermore, the analysis is extended to general GANs under
certain assumptions. We find that in order to ensure a good convergence rate,
two factors of the Jacobian in the GAN training dynamics should be
simultaneously avoided, which are (i) the Phase Factor, i.e., the Jacobian has
complex eigenvalues with a large imaginary-to-real ratio, and (ii) the
Conditioning Factor, i.e., the Jacobian is ill-conditioned. Previous methods of
regularizing the Jacobian can only alleviate one of these two factors, while
making the other more severe. Thus we propose a new JAcobian REgularization
(JARE) for GANs, which simultaneously addresses both factors by construction.
Finally, we conduct experiments that confirm our theoretical analysis and
demonstrate the advantages of JARE over previous methods in stabilizing GANs.
| stat |
Trade-offs in Large-Scale Distributed Tuplewise Estimation and Learning | The development of cluster computing frameworks has allowed practitioners to
scale out various statistical estimation and machine learning algorithms with
minimal programming effort. This is especially true for machine learning
problems whose objective function is nicely separable across individual data
points, such as classification and regression. In contrast, statistical
learning tasks involving pairs (or more generally tuples) of data points - such
as metric learning, clustering or ranking do not lend themselves as easily to
data-parallelism and in-memory computing. In this paper, we investigate how to
balance between statistical performance and computational efficiency in such
distributed tuplewise statistical problems. We first propose a simple strategy
based on occasionally repartitioning data across workers between parallel
computation stages, where the number of repartitioning steps rules the
trade-off between accuracy and runtime. We then present some theoretical
results highlighting the benefits brought by the proposed method in terms of
variance reduction, and extend our results to design distributed stochastic
gradient descent algorithms for tuplewise empirical risk minimization. Our
results are supported by numerical experiments in pairwise statistical
estimation and learning on synthetic and real-world datasets.
| stat |
Classification Algorithm for High Dimensional Protein Markers in
Time-course Data | Identification of biomarkers is an emerging area in Oncology. In this
article, we develop an efficient statistical procedure for classification of
protein markers according to their effect on cancer progression. A
high-dimensional time-course dataset of protein markers for 80 patients
motivates us for developing the model. We obtain the optimal threshold values
for markers using Cox proportional hazard model. The optimal threshold value is
defined as a level of a marker having maximum impact on cancer progression. The
classification was validated by comparing random components using both
proportional hazard and accelerated failure time frailty models. The study
elucidates the application of two separate joint modeling techniques using auto
regressive-type model and mixed effect model for time-course data and
proportional hazard model for survival data with proper utilization of Bayesian
methodology. Also, a prognostic score has been developed on the basis of few
selected genes with application on patients. The complete analysis is performed
by R programming code. This study facilitates to identify relevant biomarkers
from a set of markers.
| stat |
Efficiently Sampling Functions from Gaussian Process Posteriors | Gaussian processes are the gold standard for many real-world modeling
problems, especially in cases where a model's success hinges upon its ability
to faithfully represent predictive uncertainty. These problems typically exist
as parts of larger frameworks, wherein quantities of interest are ultimately
defined by integrating over posterior distributions. These quantities are
frequently intractable, motivating the use of Monte Carlo methods. Despite
substantial progress in scaling up Gaussian processes to large training sets,
methods for accurately generating draws from their posterior distributions
still scale cubically in the number of test locations. We identify a
decomposition of Gaussian processes that naturally lends itself to scalable
sampling by separating out the prior from the data. Building off of this
factorization, we propose an easy-to-use and general-purpose approach for fast
posterior sampling, which seamlessly pairs with sparse approximations to afford
scalability both during training and at test time. In a series of experiments
designed to test competing sampling schemes' statistical properties and
practical ramifications, we demonstrate how decoupled sample paths accurately
represent Gaussian process posteriors at a fraction of the usual cost.
| stat |
Stochastic Gradient Descent with Exponential Convergence Rates of
Expected Classification Errors | We consider stochastic gradient descent and its averaging variant for binary
classification problems in a reproducing kernel Hilbert space. In the
traditional analysis using a consistency property of loss functions, it is
known that the expected classification error converges more slowly than the
expected risk even when assuming a low-noise condition on the conditional label
probabilities. Consequently, the resulting rate is sublinear. Therefore, it is
important to consider whether much faster convergence of the expected
classification error can be achieved. In recent research, an exponential
convergence rate for stochastic gradient descent was shown under a strong
low-noise condition but provided theoretical analysis was limited to the
squared loss function, which is somewhat inadequate for binary classification
tasks. In this paper, we show an exponential convergence of the expected
classification error in the final phase of the stochastic gradient descent for
a wide class of differentiable convex loss functions under similar assumptions.
As for the averaged stochastic gradient descent, we show that the same
convergence rate holds from the early phase of training. In experiments, we
verify our analyses on the $L_2$-regularized logistic regression.
| stat |
Dynamic models using score copula innovations | This paper introduces a new class of observation driven dynamic models. The
time evolving parameters are driven by innovations of copula form. The
resulting models can be made strictly stationary and the innovation term is
typically chosen to be Gaussian. The innovations are formed by applying a
copula approach for the conditional score function which has close connections
the existing literature on GAS models. This new method provides a unified
framework for observation-driven models allowing the likelihood to be
explicitly computed using the prediction decomposition. The approach may be
used for multiple lag structures and for multivariate models. Strict
stationarity can be easily imposed upon the models making the invariant
properties simple to ascertain. This property also has advantages for
specifying the initial conditions needed for maximum likelihood estimation. One
step and multi-period forecasting is straight-forward and the forecasting
density is either in closed form or a simple mixture over a univariate
component. The approach is very general and the illustrations focus on
volatility models and duration models. We illustrate the performance of the
modelling approach for both univariate and multivariate volatility models.
| stat |
High Dimensional Bayesian Network Classification with Network
Global-Local Shrinkage Priors | This article proposes a novel Bayesian classification framework for networks
with labeled nodes. While literature on statistical modeling of network data
typically involves analysis of a single network, the recent emergence of
complex data in several biological applications, including brain imaging
studies, presents a need to devise a network classifier for subjects. This
article considers an application from a brain connectome study, where the
overarching goal is to classify subjects into two separate groups based on
their brain network data, along with identifying influential regions of
interest (ROIs) (referred to as nodes). Existing approaches either treat all
edge weights as a long vector or summarize the network information with a few
summary measures. Both these approaches ignore the full network structure, may
lead to less desirable inference in small samples and are not designed to
identify significant network nodes. We propose a novel binary logistic
regression framework with the network as the predictor and a binary response,
the network predictor coefficient being modeled using a novel class
global-local shrinkage priors. The framework is able to accurately detect nodes
and edges in the network influencing the classification. Our framework is
implemented using an efficient Markov Chain Monte Carlo algorithm.
Theoretically, we show asymptotically optimal classification for the proposed
framework when the number of network edges grows faster than the sample size.
The framework is empirically validated by extensive simulation studies and
analysis of a brain connectome data.
| stat |
Estimating spatially varying health effects in app-based citizen science
research | Wildland fire smoke exposures present an increasingly severe threat to public
health, and thus there is a growing need for studying the effects of protective
behaviors on improving health. Emerging smartphone applications provide
unprecedented opportunities to study this important problem, but also pose
novel challenges. Smoke Sense, a citizen science project, provides an
interactive platform for participants to engage with a smartphone app that
records air quality, health symptoms, and behaviors taken to reduce smoke
exposures. We propose a new, doubly robust estimator of the structural nested
mean model that accounts for spatially- and time-varying effects via a local
estimating equation approach with geographical kernel weighting. Moreover, our
analytical framework is flexible enough to handle informative missingness by
inverse probability weighting of estimating functions. We evaluate the new
method using extensive simulation studies and apply it to Smoke Sense survey
data collected from smartphones for a better understanding of the relationship
between smoke preventive measures and health effects. Our results estimate how
the protective behaviors' effects vary over space and time and find that
protective behaviors have more significant effects on reducing health symptoms
in the Southwest than the Northwest region of the USA.
| stat |
On the consistency of supervised learning with missing values | In many application settings, the data have missing entries which make
analysis challenging. An abundant literature addresses missing values in an
inferential framework: estimating parameters and their variance from incomplete
tables. Here, we consider supervised-learning settings: predicting a target
when missing values appear in both training and testing data. We show the
consistency of two approaches in prediction. A striking result is that the
widely-used method of imputing with a constant, such as the mean prior to
learning is consistent when missing values are not informative. This contrasts
with inferential settings where mean imputation is pointed at for distorting
the distribution of the data. That such a simple approach can be consistent is
important in practice. We also show that a predictor suited for complete
observations can predict optimally on incomplete data,through multiple
imputation.Finally, to compare imputation with learning directly with a model
that accounts for missing values, we analyze further decision trees. These can
naturally tackle empirical risk minimization with missing values, due to their
ability to handle the half-discrete nature of incomplete variables. After
comparing theoretically and empirically different missing values strategies in
trees, we recommend using the "missing incorporated in attribute" method as it
can handle both non-informative and informative missing values.
| stat |
Ensemble Methods for Survival Data with Time-Varying Covariates | Survival data with time-varying covariates are common in practice. If
relevant, such covariates can improve on the estimation of a survival function.
However, the traditional survival forests - conditional inference forest,
relative risk forest and random survival forest - have accommodated only
time-invariant covariates.
We generalize the conditional inference and relative risk forests to allow
time-varying covariates. We compare their performance with that of the extended
Cox model, a commonly used method, and the transformation forest method,
designed to detect non-proportional hazards deviations and adapted here to
accommodate time-varying covariates, through a comprehensive simulation study
in which the Kaplan-Meier estimate serves as a benchmark and the integrated L2
difference between the true and estimated survival functions is used for
evaluation.
In general, the performance of the two proposed forests substantially
improves over the Kaplan-Meier estimate. Under the proportional-hazard setting,
the best method is always one of the two proposed forests, while under the
non-proportional hazards setting, it is the adapted transformation forest. We
use K-fold cross-validation to choose between the methods, which is shown to be
an effective tool to provide guidance in practice. The performance of the
proposed forest methods for time-invariant covariate data is broadly similar to
that found for time-varying covariate data.
| stat |
Benchmarking the Neural Linear Model for Regression | The neural linear model is a simple adaptive Bayesian linear regression
method that has recently been used in a number of problems ranging from
Bayesian optimization to reinforcement learning. Despite its apparent successes
in these settings, to the best of our knowledge there has been no systematic
exploration of its capabilities on simple regression tasks. In this work we
characterize these on the UCI datasets, a popular benchmark for Bayesian
regression models, as well as on the recently introduced UCI "gap" datasets,
which are better tests of out-of-distribution uncertainty. We demonstrate that
the neural linear model is a simple method that shows generally good
performance on these tasks, but at the cost of requiring good hyperparameter
tuning.
| stat |
Scalable Bayesian neural networks by layer-wise input augmentation | We introduce implicit Bayesian neural networks, a simple and scalable
approach for uncertainty representation in deep learning. Standard Bayesian
approach to deep learning requires the impractical inference of the posterior
distribution over millions of parameters. Instead, we propose to induce a
distribution that captures the uncertainty over neural networks by augmenting
each layer's inputs with latent variables. We present appropriate input
distributions and demonstrate state-of-the-art performance in terms of
calibration, robustness and uncertainty characterisation over large-scale,
multi-million parameter image classification tasks.
| stat |
Causal inference with multiple versions of treatment and application to
personalized medicine | The development of high-throughput sequencing and targeted therapies has led
to the emergence of personalized medicine: a patient's molecular profile or the
presence of a specific biomarker of drug response will correspond to a
treatment recommendation made either by a physician or by a treatment
assignment algorithm. The growing number of such algorithms raises the question
of how to quantify their clinical impact knowing that a personalized medicine
strategy will inherently include different versions of treatment.
We thus specify an appropriate causal framework with multiple versions of
treatment to define the causal effects of interest for precision medicine
strategies and estimate them emulating clinical trials with observational data.
Therefore, we determine whether the treatment assignment algorithm is more
efficient than different control arms: gold standard treatment, observed
treatments or random assignment of targeted treatments.
Causal estimates of the precision medicine effects are first evaluated on
simulated data and they demonstrate a lower biases and variances compared with
naive estimation of the difference in expected outcome between treatment arms.
The various simulations scenarios also point out the different bias sources
depending on the clinical situation (heterogeneity of response, assignment of
observed treatments etc.). A RShiny interactive application is also provided to
further explore other user-defined scenarios. The method is then applied to
data from patient-derived xenografts (PDX): each patient tumour is implanted in
several immunodeficient cloned mice later treated with different drugs, thus
providing access to all corresponding drug sensitivities for all patients.
Access to these unique pre-clinical data emulating counterfactual outcomes
allows to validate the reliability of causal estimates obtained with the
proposed method.
| stat |
Non-Bayesian Social Learning with Gaussian Uncertain Models | Non-Bayesian social learning theory provides a framework for distributed
inference of a group of agents interacting over a social network by
sequentially communicating and updating beliefs about the unknown state of the
world through likelihood updates from their observations. Typically, likelihood
models are assumed known precisely. However, in many situations the models are
generated from sparse training data due to lack of data availability, high cost
of collection/calibration, limits within the communications network, and/or the
high dynamics of the operational environment. Recently, social learning theory
was extended to handle those model uncertainties for categorical models. In
this paper, we introduce the theory of Gaussian uncertain models and study the
properties of the beliefs generated by the network of agents. We show that even
with finite amounts of training data, non-Bayesian social learning can be
achieved and all agents in the network will converge to a consensus belief that
provably identifies the best estimate for the state of the world given the set
of prior information.
| stat |
A Poisson Kalman filter for disease surveillance | An optimal filter for Poisson observations is developed as a variant of the
traditional Kalman filter. Poisson distributions are characteristic of
infectious diseases, which model the number of patients recorded as presenting
each day to a health care system. We develop both a linear and nonlinear
(extended) filter. The methods are applied to a case study of neonatal sepsis
and postinfectious hydrocephalus in Africa, using parameters estimated from
publicly available data. Our approach is applicable to a broad range of disease
dynamics, including both noncommunicable and the inherent nonlinearities of
communicable infectious diseases and epidemics such as from COVID-19.
| stat |
Time Varying Markov Process with Partially Observed Aggregate Data; An
Application to Coronavirus | A major difficulty in the analysis of propagation of the coronavirus is that
many infected individuals show no symptoms of Covid-19. This implies a lack of
information on the total counts of infected individuals and of recovered and
immunized individuals. In this paper, we consider parametric time varying
Markov processes of Coronavirus propagation and show how to estimate the model
parameters and approximate the unobserved counts from daily numbers of infected
and detected individuals and total daily death counts. This model-based
approach is illustrated in an application to French data.
| stat |
Causal bounds for outcome-dependent sampling in observational studies | Outcome-dependent sampling designs are common in many different scientific
fields including epidemiology, ecology, and economics. As with all
observational studies, such designs often suffer from unmeasured confounding,
which generally precludes the nonparametric identification of causal effects.
Nonparametric bounds can provide a way to narrow the range of possible values
for a nonidentifiable causal effect without making additional untestable
assumptions. The nonparametric bounds literature has almost exclusively focused
on settings with random sampling, and the bounds have often been derived with a
particular linear programming method. We derive novel bounds for the causal
risk difference, often referred to as the average treatment effect, in six
settings with outcome-dependent sampling and unmeasured confounding for a
binary outcome and exposure. Our derivations of the bounds illustrate two
approaches that may be applicable in other settings where the bounding problem
cannot be directly stated as a system of linear constraints. We illustrate our
derived bounds in a real data example involving the effect of vitamin D
concentration on mortality.
| stat |
Calibration for computer experiments with binary responses and
application to cell adhesion study | Calibration refers to the estimation of unknown parameters which are present
in computer experiments but not available in physical experiments. An accurate
estimation of these parameters is important because it provides a scientific
understanding of the underlying system which is not available in physical
experiments. Most of the work in the literature is limited to the analysis of
continuous responses. Motivated by a study of cell adhesion experiments, we
propose a new calibration framework for binary responses. Its application to
the T cell adhesion data provides insight into the unknown values of the
kinetic parameters which are difficult to determine by physical experiments due
to the limitation of the existing experimental techniques.
| stat |
Black-box sampling for weakly smooth Langevin Monte Carlo using
p-generalized Gaussian smoothing | Discretization of continuous-time diffusion processes is a widely recognized
method for sampling. However, the canonical Euler-Maruyama discretization of
the Langevin diffusion process, also named as Langevin Monte Carlo (LMC),
studied mostly in the context of smooth (gradient-Lipschitz) and strongly
log-concave densities, a significant constraint for its deployment in many
sciences, including computational statistics and statistical learning. In this
paper, we establish several theoretical contributions to the literature on such
sampling methods. Particularly, we generalize the Gaussian smoothing,
approximate the gradient using p-generalized Gaussian smoothing and take
advantage of it in the context of black-box sampling. We first present a
non-strongly concave and weakly smooth black-box LMC algorithm, ideal for
practical applicability of sampling challenges in a general setting.
| stat |
A posteriori stochastic correction of reduced models in delayed
acceptance MCMC, with application to multiphase subsurface inverse problems | Sample-based Bayesian inference provides a route to uncertainty
quantification in the geosciences, and inverse problems in general, though is
very computationally demanding in the naive form that requires simulating an
accurate computer model at each iteration. We present a new approach that
constructs a stochastic correction to the error induced by a reduced model,
with the correction improving as the algorithm proceeds. This enables sampling
from the correct target distribution at reduced computational cost per
iteration, as in existing delayed-acceptance schemes, while avoiding
appreciable loss of statistical efficiency that necessarily occurs when using a
reduced model. Use of the stochastic correction significantly reduces the
computational cost of estimating quantities of interest within desired
uncertainty bounds. In contrast, existing schemes that use a reduced model
directly as a surrogate do not actually improve computational efficiency in our
target applications. We build on recent simplified conditions for adaptive
Markov chain Monte Carlo algorithms to give practical approximation schemes and
algorithms with guaranteed convergence. The efficacy of this new approach is
demonstrated in two computational examples, including calibration of a
large-scale numerical model of a real geothermal reservoir, that show good
computational and statistical efficiencies on both synthetic and measured data
sets.
| stat |
Data-Driven Modeling Reveals the Impact of Stay-at-Home Orders on Human
Mobility during the COVID-19 Pandemic in the U.S | One approach to delay the spread of the novel coronavirus (COVID-19) is to
reduce human travel by imposing travel restriction policies. It is yet unclear
how effective those policies are on suppressing the mobility trend due to the
lack of ground truth and large-scale dataset describing human mobility during
the pandemic. This study uses real-world location-based service data collected
from anonymized mobile devices to uncover mobility changes during COVID-19 and
under the 'Stay-at-home' state orders in the U.S. The study measures human
mobility with two important metrics: daily average number of trips per person
and daily average person-miles traveled. The data-driven analysis and modeling
attribute less than 5% of the reduction in the number of trips and person-miles
traveled to the effect of the policy. The models developed in the study exhibit
high prediction accuracy and can be applied to inform epidemics modeling with
empirically verified mobility trends and to support time-sensitive
decision-making processes.
| stat |
Regularized deep learning with nonconvex penalties | Regularization methods are often employed in deep learning neural networks
(DNNs) to prevent overfitting. For penalty based DNN regularization methods,
convex penalties are typically considered because of their optimization
guarantees. Recent theoretical work have shown that nonconvex penalties that
satisfy certain regularity conditions are also guaranteed to perform well with
standard optimization algorithms. In this paper, we examine new and currently
existing nonconvex penalties for DNN regularization. We provide theoretical
justifications for the new penalties and also assess the performance of all
penalties with DNN analyses of seven datasets.
| stat |
Sparse representation for damage identification of structural systems | Identifying damage of structural systems is typically characterized as an
inverse problem which might be ill-conditioned due to aleatory and epistemic
uncertainties induced by measurement noise and modeling error. Sparse
representation can be used to perform inverse analysis for the case of sparse
damage. In this paper, we propose a novel two-stage sensitivity analysis-based
framework for both model updating and sparse damage identification.
Specifically, an $\ell_2$ Bayesian learning method is firstly developed for
updating the intact model and uncertainty quantification so as to set forward a
baseline for damage detection. A sparse representation pipeline built on a
quasi-$\ell_0$ method, e.g., Sequential Threshold Least Squares (STLS)
regression, is then presented for damage localization and quantification.
Additionally, Bayesian optimization together with cross validation is developed
to heuristically learn hyperparameters from data, which saves the computational
cost of hyperparameter tuning and produces more reliable identification result.
The proposed framework is verified by three examples, including a 10-story
shear-type building, a complex truss structure, and a shake table test of an
eight-story steel frame. Results show that the proposed approach is capable of
both localizing and quantifying structural damage with high accuracy.
| stat |
An Introduction to Proximal Causal Learning | A standard assumption for causal inference from observational data is that
one has measured a sufficiently rich set of covariates to ensure that within
covariate strata, subjects are exchangeable across observed treatment values.
Skepticism about the exchangeability assumption in observational studies is
often warranted because it hinges on investigators' ability to accurately
measure covariates capturing all potential sources of confounding.
Realistically, confounding mechanisms can rarely if ever, be learned with
certainty from measured covariates. One can therefore only ever hope that
covariate measurements are at best proxies of true underlying confounding
mechanisms operating in an observational study, thus invalidating causal claims
made on basis of standard exchangeability conditions. Causal learning from
proxies is a challenging inverse problem which has to date remained unresolved.
In this paper, we introduce a formal potential outcome framework for proximal
causal learning, which while explicitly acknowledging covariate measurements as
imperfect proxies of confounding mechanisms, offers an opportunity to learn
about causal effects in settings where exchangeability on the basis of measured
covariates fails. Sufficient conditions for nonparametric identification are
given, leading to the proximal g-formula and corresponding proximal
g-computation algorithm for estimation. These may be viewed as generalizations
of Robins' foundational g-formula and g-computation algorithm, which account
explicitly for bias due to unmeasured confounding. Both point treatment and
time-varying treatment settings are considered, and an application of proximal
g-computation of causal effects is given for illustration.
| stat |
A Monte Carlo EM Algorithm for the Parameter Estimation of Aggregated
Hawkes Processes | A key difficulty that arises from real event data is imprecision in the
recording of event time-stamps. In many cases, retaining event times with a
high precision is expensive due to the sheer volume of activity. Combined with
practical limits on the accuracy of measurements, aggregated data is common. In
order to use point processes to model such event data, tools for handling
parameter estimation are essential. Here we consider parameter estimation of
the Hawkes process, a type of self-exciting point process that has found
application in the modeling of financial stock markets, earthquakes and social
media cascades. We develop a novel optimization approach to parameter
estimation of aggregated Hawkes processes using a Monte Carlo
Expectation-Maximization (MC-EM) algorithm. Through a detailed simulation
study, we demonstrate that existing methods are capable of producing severely
biased and highly variable parameter estimates and that our novel MC-EM method
significantly outperforms them in all studied circumstances. These results
highlight the importance of correct handling of aggregated data.
| stat |
On the Sample Complexity of Privately Learning Unbounded
High-Dimensional Gaussians | We provide sample complexity upper bounds for agnostically learning
multivariate Gaussians under the constraint of approximate differential
privacy. These are the first finite sample upper bounds for general Gaussians
which do not impose restrictions on the parameters of the distribution. Our
bounds are near-optimal in the case when the covariance is known to be the
identity, and conjectured to be near-optimal in the general case. From a
technical standpoint, we provide analytic tools for arguing the existence of
global "locally small" covers from local covers of the space. These are
exploited using modifications of recent techniques for differentially private
hypothesis selection. Our techniques may prove useful for privately learning
other distribution classes which do not possess a finite cover.
| stat |
BoMb-OT: On Batch of Mini-batches Optimal Transport | Mini-batch optimal transport (m-OT) has been successfully used in practical
applications that involve probability measures with intractable density, or
probability measures with a very high number of supports. The m-OT solves
several sparser optimal transport problems and then returns the average of
their costs and transportation plans. Despite its scalability advantage, m-OT
is not a proper metric between probability measures since it does not satisfy
the identity property. To address this problem, we propose a novel
mini-batching scheme for optimal transport, named Batch of Mini-batches Optimal
Transport (BoMb-OT), that can be formulated as a well-defined distance on the
space of probability measures. Furthermore, we show that the m-OT is a limit of
the entropic regularized version of the proposed BoMb-OT when the regularized
parameter goes to infinity. We carry out extensive experiments to show that the
new mini-batching scheme can estimate a better transportation plan between two
original measures than m-OT. It leads to a favorable performance of BoMb-OT in
the matching and color transfer tasks. Furthermore, we observe that BoMb-OT
also provides a better objective loss than m-OT for doing approximate Bayesian
computation, estimating parameters of interest in parametric generative models,
and learning non-parametric generative models with gradient flow.
| stat |
Mat\'ern Gaussian Processes on Graphs | Gaussian processes are a versatile framework for learning unknown functions
in a manner that permits one to utilize prior information about their
properties. Although many different Gaussian process models are readily
available when the input space is Euclidean, the choice is much more limited
for Gaussian processes whose input space is an undirected graph. In this work,
we leverage the stochastic partial differential equation characterization of
Mat\'ern Gaussian processes - a widely-used model class in the Euclidean
setting - to study their analog for undirected graphs. We show that the
resulting Gaussian processes inherit various attractive properties of their
Euclidean and Riemannian analogs and provide techniques that allow them to be
trained using standard methods, such as inducing points. This enables graph
Mat\'ern Gaussian processes to be employed in mini-batch and non-conjugate
settings, thereby making them more accessible to practitioners and easier to
deploy within larger learning frameworks.
| stat |
Picasso: A Sparse Learning Library for High Dimensional Data Analysis in
R and Python | We describe a new library named picasso, which implements a unified framework
of pathwise coordinate optimization for a variety of sparse learning problems
(e.g., sparse linear regression, sparse logistic regression, sparse Poisson
regression and scaled sparse linear regression) combined with efficient active
set selection strategies. Besides, the library allows users to choose different
sparsity-inducing regularizers, including the convex $\ell_1$, nonconvex MCP
and SCAD regularizers. The library is coded in C++ and has user-friendly R and
Python wrappers. Numerical experiments demonstrate that picasso can scale up to
large problems efficiently.
| stat |
FuDGE: Functional Differential Graph Estimation with fully and
discretely observed curves | We consider the problem of estimating the difference between two functional
undirected graphical models with shared structures. In many applications, data
are naturally regarded as high-dimensional random function vectors rather than
multivariate scalars. For example, electroencephalography (EEG) data are more
appropriately treated as functions of time. In these problems, not only can the
number of functions measured per sample be large, but each function is itself
an infinite dimensional object, making estimation of model parameters
challenging. In practice, curves are usually discretely observed, which makes
graph structure recovery even more challenging. We formally characterize when
two functional graphical models are comparable and propose a method that
directly estimates the functional differential graph, which we term FuDGE.
FuDGE avoids separate estimation of each graph, which allows for estimation in
problems where individual graphs are dense, but their difference is sparse. We
show that FuDGE consistently estimates the functional differential graph in a
high-dimensional setting for both discretely observed and fully observed
function paths. We illustrate finite sample properties of our method through
simulation studies. In order to demonstrate the benefits of our method, we
propose Joint Functional Graphical Lasso as a competitor, which is a
generalization of the Joint Graphical Lasso. Finally, we apply our method to
EEG data to uncover differences in functional brain connectivity between
alcoholics and control subjects.
| stat |
Efficient quantification of the impact of demand and weather uncertainty
in power system models | This paper introduces a new approach to quantify the impact of forward
propagated demand and weather uncertainty on power system planning and
operation models. Recent studies indicate that such sampling uncertainty,
originating from demand and weather time series inputs, should not be ignored.
However, established uncertainty quantification approaches fail in this context
due to the data and computing resources required for standard Monte Carlo
analysis with disjoint samples. The method introduced here uses an m out of n
bootstrap with shorter time series than the original, enhancing computational
efficiency and avoiding the need for any additional data. It both quantifies
output uncertainty and determines the sample length required for desired
confidence levels. Simulations and validation exercises are performed on two
capacity expansion planning models and one unit commitment and economic
dispatch model. A diagnostic for the validity of estimated uncertainty bounds
is discussed. The models, data and code are made available.
| stat |
Focused Bayesian Prediction | We propose a new method for conducting Bayesian prediction that delivers
accurate predictions without correctly specifying the unknown true data
generating process. A prior is defined over a class of plausible predictive
models. After observing data, we update the prior to a posterior over these
models, via a criterion that captures a user-specified measure of predictive
accuracy. Under regularity, this update yields posterior concentration onto the
element of the predictive class that maximizes the expectation of the accuracy
measure. In a series of simulation experiments and empirical examples we find
notable gains in predictive accuracy relative to conventional likelihood-based
prediction.
| stat |
DECoVaC: Design of Experiments with Controlled Variability Components | Reproducible research in Machine Learning has seen a salutary abundance of
progress lately: workflows, transparency, and statistical analysis of
validation and test performance. We build on these efforts and take them
further. We offer a principled experimental design methodology, based on linear
mixed models, to study and separate the effects of multiple factors of
variation in machine learning experiments. This approach allows to account for
the effects of architecture, optimizer, hyper-parameters, intentional
randomization, as well as unintended lack of determinism across reruns. We
illustrate that methodology by analyzing Matching Networks, Prototypical
Networks and TADAM on the miniImagenet dataset.
| stat |
Structure learning of Bayesian networks involving cyclic structures | Many biological networks include cyclic structures. In such cases, Bayesian
networks (BNs), which must be acyclic, are not sound models for structure
learning. Dynamic BNs can be used but require relatively large time series
data. We discuss an alternative model that embeds cyclic structures within
acyclic BNs, allowing us to still use the factorization property and
informative priors on network structure. We present an implementation in the
linear Gaussian case, where cyclic structures are treated as multivariate
nodes. We use a Markov Chain Monte Carlo algorithm for inference, allowing us
to work with posterior distribution on the space of graphs.
| stat |
Assortativity measures for weighted and directed networks | Assortativity measures the tendency of a vertex in a network being connected
by other vertexes with respect to some vertex-specific features. Classical
assortativity coefficients are defined for unweighted and undirected networks
with respect to vertex degree. We propose a class of assortativity coefficients
that capture the assortative characteristics and structure of weighted and
directed networks more precisely. The vertex-to-vertex strength correlation is
used as an example, but the proposed measure can be applied to any pair of
vertex-specific features. The effectiveness of the proposed measure is assessed
through extensive simulations based on prevalent random network models in
comparison with existing assortativity measures. In application World
Input-Ouput Networks,the new measures reveal interesting insights that would
not be obtained by using existing ones. An implementation is publicly available
in a R package "wdnet".
| stat |
Modeling Nonstationary and Asymmetric Multivariate Spatial Covariances
via Deformations | Multivariate spatial-statistical models are often used when modeling
environmental and socio-demographic processes. The most commonly used models
for multivariate spatial covariances assume both stationarity and symmetry for
the cross-covariances, but these assumptions are rarely tenable in practice. In
this article we introduce a new and highly flexible class of nonstationary and
asymmetric multivariate spatial covariance models that are constructed by
modeling the simpler and more familiar stationary and symmetric multivariate
covariances on a warped domain. Inspired by recent developments in the
univariate case, we propose modeling the warping function as a composition of a
number of simple injective warping functions in a deep-learning framework.
Importantly, covariance-model validity is guaranteed by construction. We
establish the types of warpings that allow for cross-covariance symmetry and
asymmetry, and we use likelihood-based methods for inference that are
computationally efficient. The utility of this new class of models is shown
through two data illustrations: a simulation study on nonstationary data and an
application on ocean temperatures at two different depths.
| stat |
Exponential random graph model parameter estimation for very large
directed networks | Exponential random graph models (ERGMs) are widely used for modeling social
networks observed at one point in time. However the computational difficulty of
ERGM parameter estimation has limited the practical application of this class
of models to relatively small networks, up to a few thousand nodes at most,
with usually only a few hundred nodes or fewer. In the case of undirected
networks, snowball sampling can be used to find ERGM parameter estimates of
larger networks via network samples, and recently published improvements in
ERGM network distribution sampling and ERGM estimation algorithms have allowed
ERGM parameter estimates of undirected networks with over one hundred thousand
nodes to be made. However the implementations of these algorithms to date have
been limited in their scalability, and also restricted to undirected networks.
Here we describe an implementation of the recently published Equilibrium
Expectation (EE) algorithm for ERGM parameter estimation of large directed
networks. We test it on some simulated networks, and demonstrate its
application to an online social network with over 1.6 million nodes.
| stat |
The Fast Loaded Dice Roller: A Near-Optimal Exact Sampler for Discrete
Probability Distributions | This paper introduces a new algorithm for the fundamental problem of
generating a random integer from a discrete probability distribution using a
source of independent and unbiased random coin flips. We prove that this
algorithm, which we call the Fast Loaded Dice Roller (FLDR), is highly
efficient in both space and time: (i) the size of the sampler is guaranteed to
be linear in the number of bits needed to encode the input distribution; and
(ii) the expected number of bits of entropy it consumes per sample is at most 6
bits more than the information-theoretically optimal rate. We present fast
implementations of the linear-time preprocessing and near-optimal sampling
algorithms using unsigned integer arithmetic. Empirical evaluations on a broad
set of probability distributions establish that FLDR is 2x-10x faster in both
preprocessing and sampling than multiple baseline algorithms, including the
widely-used alias and interval samplers. It also uses up to 10000x less space
than the information-theoretically optimal sampler, at the expense of less than
1.5x runtime overhead.
| stat |
Novel Compositional Data's Grey Model for Structurally Forecasting
Arctic Crude Oil Import | The reserve of crude oil in the Arctic area is abundant. Ice melting is
making it possible to have intermediate access to the Arctic crude oil and its
transportation. A novel compositional data's grey model is proposed in this
paper to structurally forecast Arctic crude oil import. Firstly, the general
accumulative operation sequence of multivariate compositional data is defined
according to Aitchison geometry, then obtaining the novel model with the form
of the compositional data vectors. Secondly, this paper studies the least
square parameter estimation of the model. The novel model is deduced and
selected as the time-response expression of the solution. Thirdly, this paper
infuses the novel model with traditional grey model to improve its robustness.
Differential Evolution algorithm is introduced to determine the optimal value
of the general matrix. Lastly, two validation examples are provided for
confirming the effectiveness of the novel model by comparing it with other
existing models, before being employed to forecast the crude oil import
structure in China. The results show that the novel model provides better
performance in all crude oil cases in short-term forecasting. Therefore, by
using the new model, China's development parameter is 0.5214 and Determination
Factor of the novel model is 0.5999, which means that the crude oil import
structure of China is being changed. Specifically, the amount of crude oil
imported from the Arctic area is obviously increasing in the next 6 years,
showing sufficient proof of the edge owned by the Arctic area: abundant crude
oil reserves and shortening transportation distance.
| stat |
MaxGap Bandit: Adaptive Algorithms for Approximate Ranking | This paper studies the problem of adaptively sampling from K distributions
(arms) in order to identify the largest gap between any two adjacent means. We
call this the MaxGap-bandit problem. This problem arises naturally in
approximate ranking, noisy sorting, outlier detection, and top-arm
identification in bandits. The key novelty of the MaxGap-bandit problem is that
it aims to adaptively determine the natural partitioning of the distributions
into a subset with larger means and a subset with smaller means, where the
split is determined by the largest gap rather than a pre-specified rank or
threshold. Estimating an arm's gap requires sampling its neighboring arms in
addition to itself, and this dependence results in a novel hardness parameter
that characterizes the sample complexity of the problem. We propose elimination
and UCB-style algorithms and show that they are minimax optimal. Our
experiments show that the UCB-style algorithms require 6-8x fewer samples than
non-adaptive sampling to achieve the same error.
| stat |
Anchored Bayesian Gaussian Mixture Models | Finite mixtures are a flexible modeling tool for irregularly shaped densities
and samples from heterogeneous populations. When modeling with mixtures using
an exchangeable prior on the component features, the component labels are
arbitrary and are indistinguishable in posterior analysis. This makes it
impossible to attribute any meaningful interpretation to the marginal posterior
distributions of the component features. We propose a model in which a small
number of observations are assumed to arise from some of the labeled component
densities. The resulting model is not exchangeable, allowing inference on the
component features without post-processing. Our method assigns meaning to the
component labels at the modeling stage and can be justified as a data-dependent
informative prior on the labelings. We show that our method produces
interpretable results, often (but not always) similar to those resulting from
relabeling algorithms, with the added benefit that the marginal inferences
originate directly from a well specified probability model rather than a post
hoc manipulation. We provide asymptotic results leading to practical guidelines
for model selection that are motivated by maximizing prior information about
the class labels and demonstrate our method on real and simulated data.
| stat |
Robust convex clustering: How does fusion penalty enhance robustness? | Convex clustering has gained popularity recently due to its desirable
performance in empirical studies. It involves solving a convex optimization
problem with the cost function being a squared error loss plus a fusion penalty
that encourages the estimated centroids for observations in the same cluster to
be identical. However, when data are contaminated, convex clustering with a
squared error loss will fail to identify correct cluster memberships even when
there is only one arbitrary outlier. To address this challenge, we propose a
robust convex clustering method. Theoretically, we show that the new estimator
is resistant to arbitrary outliers: it does not break down until more than half
of the observations are arbitrary outliers. In particular, we observe a new
phenomenon that the fusion penalty can help enhance robustness. Numerical
studies are performed to demonstrate the competitive performance of the
proposed method.
| stat |
Cross-classified multilevel models | Cross-classified multilevel modelling is an extension of standard multilevel
modelling for non-hierarchical data that have cross-classified structures.
Traditional multilevel models involve hierarchical data structures whereby
lower level units such as students are nested within higher level units such as
schools and where these higher level units may in turn be nested within further
groupings or clusters such as school districts, regions, and countries. With
hierarchical data structures, there is an exact nesting of each lower level
unit in one and only one higher level unit. For example, each student attends
one school, each school is located within one school district, and so on.
However, social reality is more complicated than this, and so social and
behavioural data often do not follow pure or strict hierarchies. Two types of
non-hierarchical data structures which often appear in practice are
cross-classified and multiple membership structures. In this article, we
describe cross-classified data structures and cross-classified hierarchical
linear modelling which can be used to analyse them.
| stat |
On the Current State of Research in Explaining Ensemble Performance
Using Margins | Empirical evidence shows that ensembles, such as bagging, boosting, random
and rotation forests, generally perform better in terms of their generalization
error than individual classifiers. To explain this performance, Schapire et al.
(1998) developed an upper bound on the generalization error of an ensemble
based on the margins of the training data, from which it was concluded that
larger margins should lead to lower generalization error, everything else being
equal. Many other researchers have backed this assumption and presented tighter
bounds on the generalization error based on either the margins or functions of
the margins. For instance, Shen and Li (2010) provide evidence suggesting that
the generalization error of a voting classifier might be reduced by increasing
the mean and decreasing the variance of the margins. In this article we propose
several techniques and empirically test whether the current state of research
in explaining ensemble performance holds. We evaluate the proposed methods
through experiments with real and simulated data sets.
| stat |
Trajectory Functional Boxplots | With the development of data-monitoring techniques in various fields of
science, multivariate functional data are often observed. Consequently, an
increasing number of methods have appeared to extend the general summary
statistics of multivariate functional data. However, trajectory functional
data, as an important sub-type, have not been studied very well. This article
proposes two informative exploratory tools, the trajectory functional boxplot,
and the modified simplicial band depth (MSBD) versus Wiggliness of Directional
Outlyingness (WO) plot, to visualize the centrality of trajectory functional
data. The newly defined WO index effectively measures the shape variation of
curves and hence serves as a detector for shape outliers; additionally, MSBD
provides a center-outward ranking result and works as a detector for magnitude
outliers. Using the two measures, the functional boxplot of the trajectory
reveals center-outward patterns and potential outliers using the raw curves,
whereas the MSBD-WO plot illustrates such patterns and outliers in a space
spanned by MSBD and WO. The proposed methods are validated on hurricane path
data and migration trace data recorded from two types of birds.
| stat |
A Primer on PAC-Bayesian Learning | Generalised Bayesian learning algorithms are increasingly popular in machine
learning, due to their PAC generalisation properties and flexibility. The
present paper aims at providing a self-contained survey on the resulting
PAC-Bayes framework and some of its main theoretical and algorithmic
developments.
| stat |
Calibration tests in multi-class classification: A unifying framework | In safety-critical applications a probabilistic model is usually required to
be calibrated, i.e., to capture the uncertainty of its predictions accurately.
In multi-class classification, calibration of the most confident predictions
only is often not sufficient. We propose and study calibration measures for
multi-class classification that generalize existing measures such as the
expected calibration error, the maximum calibration error, and the maximum mean
calibration error. We propose and evaluate empirically different consistent and
unbiased estimators for a specific class of measures based on matrix-valued
kernels. Importantly, these estimators can be interpreted as test statistics
associated with well-defined bounds and approximations of the p-value under the
null hypothesis that the model is calibrated, significantly improving the
interpretability of calibration measures, which otherwise lack any meaningful
unit or scale.
| stat |
Use of Cross-validation Bayes Factors to Test Equality of Two Densities | We propose a non-parametric, two-sample Bayesian test for checking whether or
not two data sets share a common distribution. The test makes use of data
splitting ideas and does not require priors for high-dimensional parameter
vectors as do other nonparametric Bayesian procedures. We provide evidence that
the new procedure provides more stable Bayes factors than do methods based on
P\'olya trees. Somewhat surprisingly, the behavior of the proposed Bayes
factors when the two distributions are the same is usually superior to that of
P\'olya tree Bayes factors. We showcase the effectiveness of the test by
proving its consistency, conducting a simulation study and applying the test to
Higgs boson data.
| stat |
Reliable training and estimation of variance networks | We propose and investigate new complementary methodologies for estimating
predictive variance networks in regression neural networks. We derive a locally
aware mini-batching scheme that result in sparse robust gradients, and show how
to make unbiased weight updates to a variance network. Further, we formulate a
heuristic for robustly fitting both the mean and variance networks post hoc.
Finally, we take inspiration from posterior Gaussian processes and propose a
network architecture with similar extrapolation properties to Gaussian
processes. The proposed methodologies are complementary, and improve upon
baseline methods individually. Experimentally, we investigate the impact on
predictive uncertainty on multiple datasets and tasks ranging from regression,
active learning and generative modeling. Experiments consistently show
significant improvements in predictive uncertainty estimation over
state-of-the-art methods across tasks and datasets.
| stat |
Bayesian equation selection on sparse data for discovery of stochastic
dynamical systems | Often the underlying system of differential equations driving a stochastic
dynamical system is assumed to be known, with inference conditioned on this
assumption. We present a Bayesian framework for discovering this system of
differential equations under assumptions that align with real-life scenarios,
including the availability of relatively sparse data. Further, we discuss
computational strategies that are critical in teasing out the important details
about the dynamical system and algorithmic innovations to solve for acute
parameter interdependence in the absence of rich data. This gives a complete
Bayesian pathway for model identification via a variable selection paradigm and
parameter estimation of the corresponding model using only the observed data.
We present detailed computations and analysis of the Lorenz-96, Lorenz-63, and
the Orstein-Uhlenbeck system using the Bayesian framework we propose.
| stat |
Probability Link Models with Symmetric Information Divergence | This paper introduces link functions for transforming one probability
distribution to another such that the Kullback-Leibler and R\'enyi divergences
between the two distributions are symmetric. Two general classes of link models
are proposed. The first model links two survival functions and is applicable to
models such as the proportional odds and change point, which are used in
survival analysis and reliability modeling. A prototype application involving
the proportional odds model demonstrates advantages of symmetric divergence
measures over asymmetric measures for assessing the efficacy of features and
for model averaging purposes. The advantages include providing unique ranks for
models and unique information weights for model averaging with one-half as much
computation requirement of asymmetric divergences. The second model links two
cumulative probability distribution functions. This model produces a
generalized location model which are continuous counterparts of the binary
probability models such as probit and logit models. Examples include the
generalized probit and logit models which have appeared in the survival
analysis literature, and a generalized Laplace model and a generalized
Student-$t$ model, which are survival time models corresponding to the
respective binary probability models. Lastly, extensions to symmetric
divergence between survival functions and conditions for copula dependence
information are presented.
| stat |
Capabilities and Limitations of Time-lagged Autoencoders for Slow Mode
Discovery in Dynamical Systems | Time-lagged autoencoders (TAEs) have been proposed as a deep learning
regression-based approach to the discovery of slow modes in dynamical systems.
However, a rigorous analysis of nonlinear TAEs remains lacking. In this work,
we discuss the capabilities and limitations of TAEs through both theoretical
and numerical analyses. Theoretically, we derive bounds for nonlinear TAE
performance in slow mode discovery and show that in general TAEs learn a
mixture of slow and maximum variance modes. Numerically, we illustrate cases
where TAEs can and cannot correctly identify the leading slowest mode in two
example systems: a 2D "Washington beltway" potential and the alanine dipeptide
molecule in explicit water. We also compare the TAE results with those obtained
using state-free reversible VAMPnets (SRVs) as a variational-based neural
network approach for slow modes discovery, and show that SRVs can correctly
discover slow modes where TAEs fail.
| stat |
Estimation of Latent Network Flows in Bike-Sharing Systems | Estimation of latent network flows is a common problem in statistical network
analysis. The typical setting is that we know the margins of the network, i.e.
in- and outdegrees, but the flows are unobserved. In this paper, we develop a
mixed regression model to estimate network flows in a bike-sharing network if
only the hourly differences of in- and outdegrees at bike stations are known.
We also include exogenous covariates such as weather conditions. Two different
parameterizations of the model are considered to estimate 1) the whole network
flow and 2) the network margins only. The estimation of the model parameters is
proposed via an iterative penalized maximum likelihood approach. This is
exemplified by modeling network flows in the Vienna Bike-Sharing Network.
Furthermore, a simulation study is conducted to show the performance of the
model. For practical purposes it is crucial to predict when and at which
station there is a lack or an excess of bikes. For this application, our model
shows to be well suited by providing quite accurate predictions.
| stat |
MML is not consistent for Neyman-Scott | Strict Minimum Message Length (SMML) is an information-theoretic statistical
inference method widely cited (but only with informal arguments) as providing
estimations that are consistent for general estimation problems. It is,
however, almost invariably intractable to compute, for which reason only
approximations of it (known as MML algorithms) are ever used in practice. Using
novel techniques that allow for the first time direct, non-approximated
analysis of SMML solutions, we investigate the Neyman-Scott estimation problem,
an oft-cited showcase for the consistency of MML, and show that even with a
natural choice of prior neither SMML nor its popular approximations are
consistent for it, thereby providing a counterexample to the general claim.
This is the first known explicit construction of an SMML solution for a
natural, high-dimensional problem.
| stat |
SIMPCA: A framework for rotating and sparsifying principal components | We propose an algorithmic framework for computing sparse components from
rotated principal components. This methodology, called SIMPCA, is useful to
replace the unreliable practice of ignoring small coefficients of rotated
components when interpreting them. The algorithm computes genuinely sparse
components by projecting rotated principal components onto subsets of
variables. The so simplified components are highly correlated with the
corresponding components. By choosing different simplification strategies
different sparse solutions can be obtained which can be used to compare
alternative interpretations of the principal components. We give some examples
of how effective simplified solutions can be achieved with SIMPCA using some
publicly available data sets.
| stat |
Inference on Heterogeneous Quantile Treatment Effects via Rank-Score
Balancing | Understanding treatment effect heterogeneity in observational studies is of
great practical importance to many scientific fields because the same treatment
may affect different individuals differently. Quantile regression provides a
natural framework for modeling such heterogeneity. In this paper, we propose a
new method for inference on heterogeneous quantile treatment effects that
incorporates high-dimensional covariates. Our estimator combines a debiased
$\ell_1$-penalized regression adjustment with a quantile-specific covariate
balancing scheme. We present a comprehensive study of the theoretical
properties of this estimator, including weak convergence of the heterogeneous
quantile treatment effect process to the sum of two independent, centered
Gaussian processes. We illustrate the finite-sample performance of our approach
through Monte Carlo experiments and an empirical example, dealing with the
differential effect of mothers' education on infant birth weights.
| stat |
Conformal Prediction Intervals for Neural Networks Using Cross
Validation | Neural networks are among the most powerful nonlinear models used to address
supervised learning problems. Similar to most machine learning algorithms,
neural networks produce point predictions and do not provide any prediction
interval which includes an unobserved response value with a specified
probability. In this paper, we proposed the $k$-fold prediction interval method
to construct prediction intervals for neural networks based on $k$-fold cross
validation. Simulation studies and analysis of 10 real datasets are used to
compare the finite-sample properties of the prediction intervals produced by
the proposed method and the split conformal (SC) method. The results suggest
that the proposed method tends to produce narrower prediction intervals
compared to the SC method while maintaining the same coverage probability. Our
experimental results also reveal that the proposed $k$-fold prediction interval
method produces effective prediction intervals and is especially advantageous
relative to competing approaches when the number of training observations is
limited.
| stat |
PANDA: AdaPtive Noisy Data Augmentation for Regularization of Undirected
Graphical Models | We propose an AdaPtive Noise Augmentation (PANDA) technique to regularize the
estimation and construction of undirected graphical models. PANDA iteratively
optimizes the objective function given the noise augmented data until
convergence to achieve regularization on model parameters. The augmented noises
can be designed to achieve various regularization effects on graph estimation,
such as the bridge (including lasso and ridge), elastic net, adaptive lasso,
and SCAD penalization; it also realizes the group lasso and fused ridge. We
examine the tail bound of the noise-augmented loss function and establish that
the noise-augmented loss function and its minimizer converge almost surely to
the expected penalized loss function and its minimizer, respectively. We derive
the asymptotic distributions for the regularized parameters through PANDA in
generalized linear models, based on which, inferences for the parameters can be
obtained simultaneously with variable selection. We show the non-inferior
performance of PANDA in constructing graphs of different types in simulation
studies and apply PANDA to an autism spectrum disorder data to construct a
mixed-node graph. We also show that the inferences based on the asymptotic
distribution of regularized parameter estimates via PANDA achieve nominal or
near-nominal coverage and are far more efficient, compared to some existing
post-selection procedures. Computationally, PANDA can be easily programmed in
software that implements (GLMs) without resorting to complicated optimization
techniques.
| stat |
Testing for Outliers with Conformal p-values | This paper studies the construction of p-values for nonparametric outlier
detection, taking a multiple-testing perspective. The goal is to test whether
new independent samples belong to the same distribution as a reference data set
or are outliers. We propose a solution based on conformal inference, a broadly
applicable framework which yields p-values that are marginally valid but
mutually dependent for different test points. We prove these p-values are
positively dependent and enable exact false discovery rate control, although in
a relatively weak marginal sense. We then introduce a new method to compute
p-values that are both valid conditionally on the training data and independent
of each other for different test points; this paves the way to stronger type-I
error guarantees. Our results depart from classical conformal inference as we
leverage concentration inequalities rather than combinatorial arguments to
establish our finite-sample guarantees. Furthermore, our techniques also yield
a uniform confidence bound for the false positive rate of any outlier detection
algorithm, as a function of the threshold applied to its raw statistics.
Finally, the relevance of our results is demonstrated by numerical experiments
on real and simulated data.
| stat |
Bagged filters for partially observed interacting systems | Bagging (i.e., bootstrap aggregating) involves combining an ensemble of
bootstrap estimators. We consider bagging for inference from noisy or
incomplete measurements on a collection of interacting stochastic dynamic
systems. Each system is called a unit, and each unit is associated with a
spatial location. A motivating example arises in epidemiology, where each unit
is a city: the majority of transmission occurs within a city, with smaller yet
epidemiologically important interactions arising from disease transmission
between cities. Monte~Carlo filtering methods used for inference on nonlinear
non-Gaussian systems can suffer from a curse of dimensionality as the number of
units increases. We introduce bagged filter (BF) methodology which combines an
ensemble of Monte Carlo filters, using spatiotemporally localized weights to
select successful filters at each unit and time. We obtain conditions under
which likelihood evaluation using a BF algorithm can beat a curse of
dimensionality, and we demonstrate applicability even when these conditions do
not hold. BF can out-perform an ensemble Kalman filter on a coupled population
dynamics model describing infectious disease transmission. A block particle
filter also performs well on this task, though the bagged filter respects
smoothness and conservation laws that a block particle filter can violate.
| stat |
Three-dimensional Radial Visualization of High-dimensional Datasets with
Mixed Features | We develop methodology for 3D radial visualization (RadViz) of
high-dimensional datasets. Our display engine is called RadViz3D and extends
the classical 2D RadViz that visualizes multivariate data in the 2D plane by
mapping every record to a point inside the unit circle. We show that
distributing anchor points at least approximately uniformly on the 3D unit
sphere provides a better visualization with minimal artificial visual
correlation for data with uncorrelated variables. Our RadViz3D methodology
therefore places equi-spaced anchor points, one for every feature, exactly for
the five Platonic solids, and approximately via a Fibonacci grid for the other
cases. Our Max-Ratio Projection (MRP) method then utilizes the group
information in high dimensions to provide distinctive lower-dimensional
projections that are then displayed using Radviz3D. Our methodology is extended
to datasets with discrete and continuous features where a Gaussianized
distributional transform is used in conjunction with copula models before
applying MRP and visualizing the result using RadViz3D. A R package radviz3d
implementing our complete methodology is available.
| stat |
Comparing statistical methods to predict leptospirosis incidence using
hydro-climatic covariables | Leptospiroris, the infectious disease caused by the spirochete bacteria
Leptospira interrogans, constitutes an important public health problem all over
the world. In Argentina, some regions present climate and geographic
characteristics that favors the habitat of the bacteria Leptospira, whose
survival strongly depends on climatic factors. For this reason, regional public
health systems should include, as a main factor, the incidence of the disease
in order to improve the prediction of potential outbreaks, helping to stop or
delay the virus transmission. The classic methods used to perform this kind of
predictions are based in autoregressive time series tools which, as it is well
known, perform poorly when the data do not meet their requirements. Recently,
several nonparametric methods have been introduced to deal with those problems.
In this work, we compare a semiparametric method, called Semi-Functional
Partial Linear Regression (SFPLR) with the classic ARIMA and a new alternative
ARIMAX, in order to select the best predictive tool for the incidence of
leptospirosis in the Argentinian Litoral region. In particular, SFPLR and
ARIMAX are methods that allow the use of (hydrometeorological) covariables
which could improve the prediction of outbreaks of leptospirosis.
| stat |
On Testing for Biases in Peer Review | We consider the issue of biases in scholarly research, specifically, in peer
review. There is a long standing debate on whether exposing author identities
to reviewers induces biases against certain groups, and our focus is on
designing tests to detect the presence of such biases. Our starting point is a
remarkable recent work by Tomkins, Zhang and Heavlin which conducted a
controlled, large-scale experiment to investigate existence of biases in the
peer reviewing of the WSDM conference. We present two sets of results in this
paper. The first set of results is negative, and pertains to the statistical
tests and the experimental setup used in the work of Tomkins et al. We show
that the test employed therein does not guarantee control over false alarm
probability and under correlations between relevant variables coupled with any
of the following conditions, with high probability, can declare a presence of
bias when it is in fact absent: (a) measurement error, (b) model mismatch, (c)
reviewer calibration. Moreover, we show that the setup of their experiment may
itself inflate false alarm probability if (d) bidding is performed in non-blind
manner or (e) popular reviewer assignment procedure is employed. Our second set
of results is positive and is built around a novel approach to testing for
biases that we propose. We present a general framework for testing for biases
in (single vs. double blind) peer review. We then design hypothesis tests that
under minimal assumptions guarantee control over false alarm probability and
non-trivial power even under conditions (a)--(c) as well as propose an
alternative experimental setup which mitigates issues (d) and (e). Finally, we
show that no statistical test can improve over the non-parametric tests we
consider in terms of the assumptions required to control for the false alarm
probability.
| stat |
Ensemble updating of binary state vectors by maximising the expected
number of unchanged components | In recent years, several ensemble-based filtering methods have been proposed
and studied. The main challenge in such procedures is the updating of a prior
ensemble to a posterior ensemble at every step of the filtering recursions. In
the famous ensemble Kalman filter, the assumption of a linear-Gaussian state
space model is introduced in order to overcome this issue, and the prior
ensemble is updated with a linear shift closely related to the traditional
Kalman filter equations. In the current article, we consider how the ideas
underlying the ensemble Kalman filter can be applied when the components of the
state vectors are binary variables. While the ensemble Kalman filter relies on
Gaussian approximations of the forecast and filtering distributions, we instead
use first order Markov chains. To update the prior ensemble, we simulate
samples from a distribution constructed such that the expected number of equal
components in a prior and posterior state vector is maximised. We demonstrate
the performance of our approach in a simulation example inspired by the
movement of oil and water in a petroleum reservoir, where also a more na\"{i}ve
updating approach is applied for comparison. Here, we observe that the
Frobenius norm of the difference between the estimated and the true marginal
filtering probabilities is reduced to the half with our method compared to the
na\"{i}ve approach, indicating that our method is superior. Finally, we discuss
how our methodology can be generalised from the binary setting to more
complicated situations.
| stat |
Highly Scalable Bayesian Geostatistical Modeling via Meshed Gaussian
Processes on Partitioned Domains | We introduce a class of scalable Bayesian hierarchical models for the
analysis of massive geostatistical datasets. The underlying idea combines ideas
on high-dimensional geostatistics by partitioning the spatial domain and
modeling the regions in the partition using a sparsity-inducing directed
acyclic graph (DAG). We extend the model over the DAG to a well-defined spatial
process, which we call the Meshed Gaussian Process (MGP). A major contribution
is the development of a MGPs on tessellated domains, accompanied by a Gibbs
sampler for the efficient recovery of spatial random effects. In particular,
the cubic MGP (Q-MGP) can harness high-performance computing resources by
executing all large-scale operations in parallel within the Gibbs sampler,
improving mixing and computing time compared to sequential updating schemes.
Unlike some existing models for large spatial data, a Q-MGP facilitates massive
caching of expensive matrix operations, making it particularly apt in dealing
with spatiotemporal remote-sensing data. We compare Q-MGPs with large synthetic
and real world data against state-of-the-art methods. We also illustrate using
Normalized Difference Vegetation Index (NDVI) data from the Serengeti park
region to recover latent multivariate spatiotemporal random effects at millions
of locations. The source code is available at https://github.com/mkln/meshgp.
| stat |
Mehler's Formula, Branching Process, and Compositional Kernels of Deep
Neural Networks | We utilize a connection between compositional kernels and branching processes
via Mehler's formula to study deep neural networks. This new probabilistic
insight provides us a novel perspective on the mathematical role of activation
functions in compositional neural networks. We study the unscaled and rescaled
limits of the compositional kernels and explore the different phases of the
limiting behavior, as the compositional depth increases. We investigate the
memorization capacity of the compositional kernels and neural networks by
characterizing the interplay among compositional depth, sample size,
dimensionality, and non-linearity of the activation. Explicit formulas on the
eigenvalues of the compositional kernel are provided, which quantify the
complexity of the corresponding reproducing kernel Hilbert space. On the
methodological front, we propose a new random features algorithm, which
compresses the compositional layers by devising a new activation function.
| stat |
Sketching for Two-Stage Least Squares Estimation | When there is so much data that they become a computation burden, it is not
uncommon to compute quantities of interest using a sketch of data of size $m$
instead of the full sample of size $n$. This paper investigates the
implications for two-stage least squares (2SLS) estimation when the sketches
are obtained by a computationally efficient method known as CountSketch. We
obtain three results. First, we establish conditions under which given the full
sample, a sketched 2SLS estimate can be arbitrarily close to the full-sample
2SLS estimate with high probability. Second, we give conditions under which the
sketched 2SLS estimator converges in probability to the true parameter at a
rate of $m^{-1/2}$ and is asymptotically normal. Third, we show that the
asymptotic variance can be consistently estimated using the sketched sample and
suggest methods for determining an inference-conscious sketch size $m$. The
sketched 2SLS estimator is used to estimate returns to education.
| stat |
BayesSUR: An R package for high-dimensional multivariate Bayesian
variable and covariance selection in linear regression | In molecular biology, advances in high-throughput technologies have made it
possible to study complex multivariate phenotypes and their simultaneous
associations with high-dimensional genomic and other omics data, a problem that
can be studied with high-dimensional multi-response regression, where the
response variables are potentially highly correlated. To this purpose, we
recently introduced several multivariate Bayesian variable and covariance
selection models, e.g., Bayesian estimation methods for sparse seemingly
unrelated regression for variable and covariance selection. Several variable
selection priors have been implemented in this context, in particular the
hotspot detection prior for latent variable inclusion indicators, which results
in sparse variable selection for associations between predictors and multiple
phenotypes. We also propose an alternative, which uses a Markov random field
(MRF) prior for incorporating prior knowledge about the dependence structure of
the inclusion indicators. Inference of Bayesian seemingly unrelated regression
(SUR) by Markov chain Monte Carlo methods is made computationally feasible by
factorisation of the covariance matrix amongst the response variables. In this
paper we present BayesSUR, an R package, which allows the user to easily
specify and run a range of different Bayesian SUR models, which have been
implemented in C++ for computational efficiency. The R package allows the
specification of the models in a modular way, where the user chooses the priors
for variable selection and for covariance selection separately. We demonstrate
the performance of sparse SUR models with the hotspot prior and spike-and-slab
MRF prior on synthetic and real data sets representing eQTL or mQTL studies and
in vitro anti-cancer drug screening studies as examples for typical
applications.
| stat |
Prediction of Personal Protective Equipment Use in Hospitals During
COVID-19 | Demand for Personal Protective Equipment (PPE) such as surgical masks,
gloves, and gowns has increased significantly since the onset of the COVID-19
pandemic. In hospital settings, both medical staff and patients are required to
wear PPE. As these facilities resume regular operations, staff will be required
to wear PPE at all times while additional PPE will be mandated during medical
procedures. This will put increased pressure on hospitals which have had
problems predicting PPE usage and sourcing its supply. To meet this challenge,
we propose an approach to predict demand for PPE. Specifically, we model the
admission of patients to a medical department using multiple independent
queues. Each queue represents a class of patients with similar treatment plans
and hospital length-of-stay. By estimating the total workload of each class, we
derive closed-form estimates for the expected amount of PPE required over a
specified time horizon using current PPE guidelines. We apply our approach to a
data set of 22,039 patients admitted to the general internal medicine
department at St. Michael's hospital in Toronto, Canada from April 2010 to
November 2019. We find that gloves and surgical masks represent approximately
90% of predicted PPE usage. We also find that while demand for gloves is driven
entirely by patient-practitioner interactions, 86% of the predicted demand for
surgical masks can be attributed to the requirement that medical practitioners
will need to wear them when not interacting with patients.
| stat |
Perturbative estimation of stochastic gradients | In this paper we introduce a family of stochastic gradient estimation
techniques based of the perturbative expansion around the mean of the sampling
distribution. We characterize the bias and variance of the resulting
Taylor-corrected estimators using the Lagrange error formula. Furthermore, we
introduce a family of variance reduction techniques that can be applied to
other gradient estimators. Finally, we show that these new perturbative methods
can be extended to discrete functions using analytic continuation. Using this
technique, we derive a new gradient descent method for training stochastic
networks with binary weights. In our experiments, we show that the perturbative
correction improves the convergence of stochastic variational inference both in
the continuous and in the discrete case.
| stat |
A goodness of fit test for two component two parameter Weibull mixtures | Fitting mixture distributions is needed in applications where data belongs to
inhomogeneous populations comprising homogeneous sub-populations. The mixing
proportions of the sub populations are in general unknown and need to be
estimated as well. A goodness of fit test based on the empirical distribution
function is proposed for assessing the goodness of fit in model fits comprising
two components, each distributed as two parameter Weibull. The applicability of
the proposed test procedure was empirically established using a Monte Carlo
simulation study. The proposed test procedure can be easily altered to handle
two component mixtures with different component distributions.
| stat |
fairmodels: A Flexible Tool For Bias Detection, Visualization, And
Mitigation | Machine learning decision systems are getting omnipresent in our lives. From
dating apps to rating loan seekers, algorithms affect both our well-being and
future. Typically, however, these systems are not infallible. Moreover, complex
predictive models are really eager to learn social biases present in historical
data that can lead to increasing discrimination. If we want to create models
responsibly then we need tools for in-depth validation of models also from the
perspective of potential discrimination. This article introduces an R package
fairmodels that helps to validate fairness and eliminate bias in classification
models in an easy and flexible fashion. The fairmodels package offers a
model-agnostic approach to bias detection, visualization and mitigation. The
implemented set of functions and fairness metrics enables model fairness
validation from different perspectives. The package includes a series of
methods for bias mitigation that aim to diminish the discrimination in the
model. The package is designed not only to examine a single model, but also to
facilitate comparisons between multiple models.
| stat |
n-MeRCI: A new Metric to Evaluate the Correlation Between Predictive
Uncertainty and True Error | As deep learning applications are becoming more and more pervasive in
robotics, the question of evaluating the reliability of inferences becomes a
central question in the robotics community. This domain, known as predictive
uncertainty, has come under the scrutiny of research groups developing Bayesian
approaches adapted to deep learning such as Monte Carlo Dropout. Unfortunately,
for the time being, the real goal of predictive uncertainty has been swept
under the rug. Indeed, these approaches are solely evaluated in terms of raw
performance of the network prediction, while the quality of their estimated
uncertainty is not assessed. Evaluating such uncertainty prediction quality is
especially important in robotics, as actions shall depend on the confidence in
perceived information. In this context, the main contribution of this article
is to propose a novel metric that is adapted to the evaluation of relative
uncertainty assessment and directly applicable to regression with deep neural
networks. To experimentally validate this metric, we evaluate it on a toy
dataset and then apply it to the task of monocular depth estimation.
| stat |
Multilevel Monte Carlo estimation of expected information gains | The expected information gain is an important quality criterion of Bayesian
experimental designs, which measures how much the information entropy about
uncertain quantity of interest $\theta$ is reduced on average by collecting
relevant data $Y$. However, estimating the expected information gain has been
considered computationally challenging since it is defined as a nested
expectation with an outer expectation with respect to $Y$ and an inner
expectation with respect to $\theta$. In fact, the standard, nested Monte Carlo
method requires a total computational cost of $O(\varepsilon^{-3})$ to achieve
a root-mean-square accuracy of $\varepsilon$. In this paper we develop an
efficient algorithm to estimate the expected information gain by applying a
multilevel Monte Carlo (MLMC) method. To be precise, we introduce an antithetic
MLMC estimator for the expected information gain and provide a sufficient
condition on the data model under which the antithetic property of the MLMC
estimator is well exploited such that optimal complexity of
$O(\varepsilon^{-2})$ is achieved. Furthermore, we discuss how to incorporate
importance sampling techniques within the MLMC estimator to avoid arithmetic
underflow. Numerical experiments show the considerable computational cost
savings compared to the nested Monte Carlo method for a simple test case and a
more realistic pharmacokinetic model.
| stat |
Model-free posterior inference on the area under the receiver operating
characteristic curve | The area under the receiver operating characteristic curve (AUC) serves as a
summary of a binary classifier's performance. Methods for estimating the AUC
have been developed under a binormality assumption which restricts the
distribution of the score produced by the classifier. However, this assumption
introduces an infinite-dimensional nuisance parameter and can be inappropriate,
especially in the context of machine learning. This motivates us to adopt a
model-free Gibbs posterior distribution for the AUC. We present the asymptotic
Gibbs posterior concentration rate, and a strategy for tuning the learning rate
so that the corresponding credible intervals achieve the nominal frequentist
coverage probability. Simulation experiments and a real data analysis
demonstrate the Gibbs posterior's strong performance compared to existing
methods based on a rank likelihood.
| stat |
Estimating Dynamic Conditional Spread Densities to Optimise Daily
Storage Trading of Electricity | This paper formulates dynamic density functions, based upon skewed-t and
similar representations, to model and forecast electricity price spreads
between different hours of the day. This supports an optimal day ahead storage
and discharge schedule, and thereby facilitates a bidding strategy for a
merchant arbitrage facility into the day-ahead auctions for wholesale
electricity. The four latent moments of the density functions are dynamic and
conditional upon exogenous drivers, thereby permitting the mean, variance,
skewness and kurtosis of the densities to respond hourly to such factors as
weather and demand forecasts. The best specification for each spread is
selected based on the Pinball Loss function, following the closed form
analytical solutions of the cumulative density functions. Those analytical
properties also allow the calculation of risk associated with the spread
arbitrages. From these spread densities, the optimal daily operation of a
battery storage facility is determined.
| stat |
Average causal effect estimation via instrumental variables: the no
simultaneous heterogeneity assumption | Instrumental variables (IVs) can be used to provide evidence as to whether a
treatment X has a causal effect on Y. Z is a valid instrument if it satisfies
the three core IV assumptions of relevance, independence and the exclusion
restriction. Even if the instrument satisfies these assumptions, further
assumptions are required to estimate the average causal effect (ACE) of X on Y.
Sufficient assumptions for this include: homogeneity in the causal effect of X
on Y; homogeneity in the association of Z with X; and independence between X
and the causal effect of X on Y. Other assumptions allow identification of
other causal estimands. For example, the monotonicity assumption allows
identifying the average causal effect among compliers. Here, we describe the NO
Simultaneous Heterogeneity (NOSH) assumption, which requires the heterogeneity
in the Z-X association and heterogeneity in the X-Y causal effect to be
independent. We describe the necessary conditions for NOSH and show that, if
NOSH holds, conventional IV methods are consistent for the ACE even if both
homogeneity assumptions and NEM are violated. We illustrate these ideas using
simulations and by re-examining selected published studies.
| stat |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.