title
stringlengths 6
244
| abstract
stringlengths 19
5.09k
| label
stringclasses 10
values |
|---|---|---|
Efficient Clustering for Stretched Mixtures: Landscape and Optimality | This paper considers a canonical clustering problem where one receives
unlabeled samples drawn from a balanced mixture of two elliptical distributions
and aims for a classifier to estimate the labels. Many popular methods
including PCA and k-means require individual components of the mixture to be
somewhat spherical, and perform poorly when they are stretched. To overcome
this issue, we propose a non-convex program seeking for an affine transform to
turn the data into a one-dimensional point cloud concentrating around -1 and 1,
after which clustering becomes easy. Our theoretical contributions are
two-fold: (1) we show that the non-convex loss function exhibits desirable
landscape properties as long as the sample size exceeds some constant multiple
of the dimension, and (2) we leverage this to prove that an efficient
first-order algorithm achieves near-optimal statistical precision even without
good initialization. We also propose a general methodology for multi-class
clustering tasks with flexible choices of feature transforms and loss
objectives.
| stat |
Computational Optimal Transport | Optimal transport (OT) theory can be informally described using the words of
the French mathematician Gaspard Monge (1746-1818): A worker with a shovel in
hand has to move a large pile of sand lying on a construction site. The goal of
the worker is to erect with all that sand a target pile with a prescribed shape
(for example, that of a giant sand castle). Naturally, the worker wishes to
minimize her total effort, quantified for instance as the total distance or
time spent carrying shovelfuls of sand. Mathematicians interested in OT cast
that problem as that of comparing two probability distributions, two different
piles of sand of the same volume. They consider all of the many possible ways
to morph, transport or reshape the first pile into the second, and associate a
"global" cost to every such transport, using the "local" consideration of how
much it costs to move a grain of sand from one place to another. Recent years
have witnessed the spread of OT in several fields, thanks to the emergence of
approximate solvers that can scale to sizes and dimensions that are relevant to
data sciences. Thanks to this newfound scalability, OT is being increasingly
used to unlock various problems in imaging sciences (such as color or texture
processing), computer vision and graphics (for shape manipulation) or machine
learning (for regression, classification and density fitting). This short book
reviews OT with a bias toward numerical methods and their applications in data
sciences, and sheds lights on the theoretical properties of OT that make it
particularly useful for some of these applications.
| stat |
Mean and Covariance Estimation for Functional Snippets | We consider estimation of mean and covariance functions of functional
snippets, which are short segments of functions possibly observed irregularly
on an individual specific subinterval that is much shorter than the entire
study interval. Estimation of the covariance function for functional snippets
is challenging since information for the far off-diagonal regions of the
covariance structure is completely missing. We address this difficulty by
decomposing the covariance function into a variance function component and a
correlation function component. The variance function can be effectively
estimated nonparametrically, while the correlation part is modeled
parametrically, possibly with an increasing number of parameters, to handle the
missing information in the far off-diagonal regions. Both theoretical analysis
and numerical simulations suggest that this hybrid strategy %
divide-and-conquer strategy is effective. In addition, we propose a new
estimator for the variance of measurement errors and analyze its asymptotic
properties. This estimator is required for the estimation of the variance
function from noisy measurements.
| stat |
Estimating COVID-19 cases and reproduction number in Mexico | In this report we fit a semi-mechanistic Bayesian hierarchical model to
describe the Mexican COVID-19 epidemic. We obtain two epidemiological measures:
the number of infections and the reproduction number. Estimations are based on
death data. Hence, we expect our estimates to be more accurate than the attack
rates estimated from the reported number of cases.
| stat |
A Linear Mixed Model Formulation for Spatio-Temporal Random Processes
with Computational Advances for the Separable and Product-Sum Covariances | We describe spatio-temporal random processes using linear mixed models. We
show how many commonly used models can be viewed as special cases of this
general framework and pay close attention to models with separable or
product-sum covariances. The proposed linear mixed model formulation
facilitates the implementation of a novel algorithm using Stegle
eigendecompositions, a recursive application of the Sherman-Morrison-Woodbury
formula, and Helmert-Wolf blocking to efficiently invert separable and
product-sum covariance matrices, even when every spatial location is not
observed at every time point. We show our algorithm provides noticeable
improvements over the standard Cholesky decomposition approach. Via
simulations, we assess the performance of the separable and product-sum
covariances and identify scenarios where separable covariances are noticeably
inferior to product-sum covariances. We also compare likelihood-based and
semivariogram-based estimation and discuss benefits and drawbacks of both. We
use the proposed approach to analyze daily maximum temperature data in Oregon,
USA, during the 2019 summer. We end by offering guidelines for choosing among
these covariances and estimation methods based on properties of observed data.
| stat |
Non-convex Learning via Replica Exchange Stochastic Gradient MCMC | Replica exchange Monte Carlo (reMC), also known as parallel tempering, is an
important technique for accelerating the convergence of the conventional Markov
Chain Monte Carlo (MCMC) algorithms. However, such a method requires the
evaluation of the energy function based on the full dataset and is not scalable
to big data. The na\"ive implementation of reMC in mini-batch settings
introduces large biases, which cannot be directly extended to the stochastic
gradient MCMC (SGMCMC), the standard sampling method for simulating from deep
neural networks (DNNs). In this paper, we propose an adaptive replica exchange
SGMCMC (reSGMCMC) to automatically correct the bias and study the corresponding
properties. The analysis implies an acceleration-accuracy trade-off in the
numerical discretization of a Markov jump process in a stochastic environment.
Empirically, we test the algorithm through extensive experiments on various
setups and obtain the state-of-the-art results on CIFAR10, CIFAR100, and SVHN
in both supervised learning and semi-supervised learning tasks.
| stat |
The Full Bayesian Significance Test and the e-value -- Foundations,
theory and application in the cognitive sciences | Hypothesis testing is a central statistical method in psychological research
and the cognitive sciences. While the problems of null hypothesis significance
testing (NHST) have been debated widely, few attractive alternatives exist. In
this paper, we provide a tutorial on the Full Bayesian Significance Test (FBST)
and the e-value, which is a fully Bayesian alternative to traditional
significance tests which rely on p-values. The FBST is an advanced
methodological procedure which can be applied to several areas. In this
tutorial, we showcase with two examples of widely used statistical methods in
psychological research how the FBST can be used in practice, provide
researchers with explicit guidelines on how to conduct it and make available
R-code to reproduce all results. The FBST is an innovative method which has
clearly demonstrated to perform better than frequentist significance testing.
However, to our best knowledge, it has not been used so far in the
psychological sciences and should be of wide interest to a broad range of
researchers in psychology and the cognitive sciences.
| stat |
Prediction with Missing Data | Missing information is inevitable in real-world data sets. While imputation
is well-suited and theoretically sound for statistical inference, its relevance
and practical implementation for out-of-sample prediction remains unsettled. We
provide a theoretical analysis of widely used data imputation methods and
highlight their key deficiencies in making accurate predictions. Alternatively,
we propose adaptive linear regression, a new class of models that can be
directly trained and evaluated on partially observed data, adapting to the set
of available features. In particular, we show that certain adaptive regression
models are equivalent to impute-then-regress methods where the imputation and
the regression models are learned simultaneously instead of sequentially. We
validate our theoretical findings and adaptive regression approach with
numerical results with real-world data sets.
| stat |
How Good is the Bayes Posterior in Deep Neural Networks Really? | During the past five years the Bayesian deep learning community has developed
increasingly accurate and efficient approximate inference procedures that allow
for Bayesian inference in deep neural networks. However, despite this
algorithmic progress and the promise of improved uncertainty quantification and
sample efficiency there are---as of early 2020---no publicized deployments of
Bayesian neural networks in industrial practice. In this work we cast doubt on
the current understanding of Bayes posteriors in popular deep neural networks:
we demonstrate through careful MCMC sampling that the posterior predictive
induced by the Bayes posterior yields systematically worse predictions compared
to simpler methods including point estimates obtained from SGD. Furthermore, we
demonstrate that predictive performance is improved significantly through the
use of a "cold posterior" that overcounts evidence. Such cold posteriors
sharply deviate from the Bayesian paradigm but are commonly used as heuristic
in Bayesian deep learning papers. We put forward several hypotheses that could
explain cold posteriors and evaluate the hypotheses through experiments. Our
work questions the goal of accurate posterior approximations in Bayesian deep
learning: If the true Bayes posterior is poor, what is the use of more accurate
approximations? Instead, we argue that it is timely to focus on understanding
the origin of the improved performance of cold posteriors.
| stat |
Biconvex Clustering | Convex clustering has recently garnered increasing interest due to its
attractive theoretical and computational properties, but its merits become
limited in the face of high-dimensional data. In such settings, pairwise
affinity terms that rely on $k$-nearest neighbors become poorly specified and
Euclidean measures of fit provide weaker discriminating power. To surmount
these issues, we propose to modify the convex clustering objective so that
feature weights are optimized jointly with the centroids. The resulting problem
becomes biconvex, and as such remains well-behaved statistically and
algorithmically. In particular, we derive a fast algorithm with closed form
updates and convergence guarantees, and establish finite-sample bounds on its
prediction error. Under interpretable regularity conditions, the error bound
analysis implies consistency of the proposed estimator. Biconvex clustering
performs feature selection throughout the clustering task: as the learned
weights change the effective feature representation, pairwise affinities can be
updated adaptively across iterations rather than precomputed within a dubious
feature space. We validate the contributions on real and simulated data,
showing that our method effectively addresses the challenges of dimensionality
while reducing dependence on carefully tuned heuristics typical of existing
approaches.
| stat |
Learning Temporal Attention in Dynamic Graphs with Bilinear Interactions | Reasoning about graphs evolving over time is a challenging concept in many
domains, such as bioinformatics, physics, and social networks. We consider a
common case in which edges can be short term interactions (e.g., messaging) or
long term structural connections (e.g., friendship). In practice, long term
edges are often specified by humans. Human-specified edges can be both
expensive to produce and suboptimal for the downstream task. To alleviate these
issues, we propose a model based on temporal point processes and variational
autoencoders that learns to infer temporal attention between nodes by observing
node communication. As temporal attention drives between-node feature
propagation, using the dynamics of node interactions to learn this key
component provides more flexibility while simultaneously avoiding issues
associated with human-specified edges. We also propose a bilinear
transformation layer for pairs of node features instead of concatenation,
typically used in prior work, and demonstrate its superior performance in all
cases. In experiments on two datasets in the dynamic link prediction task, our
model often outperforms the baseline model that requires a human-specified
graph. Moreover, our learned attention is semantically interpretable and infers
connections similar to actual graphs.
| stat |
U.S. Power Resilience for 2002--2017 | Prolonged power outages debilitate the economy and threaten public health.
Existing research is generally limited in its scope to a single event, an
outage cause, or a region. Here, we provide one of the most comprehensive
analyses of U.S. power outages for 2002--2017. We categorized all outage data
collected under U.S. federal mandates into four outage causes and computed
industry-standard reliability metrics. Our spatiotemporal analysis reveals six
of the most resilient U.S. states since 2010, improvement of power resilience
against natural hazards in the south and northeast regions, and a
disproportionately large number of human attacks for its population in the
Western Electricity Coordinating Council region. Our regression analysis
identifies several statistically significant predictors and hypotheses for
power resilience. Furthermore, we propose a novel framework for analyzing
outage data using differential weighting and influential points to better
understand power resilience. We share curated data and code as Supplementary
Materials.
| stat |
An Empirical Bayes Method for Chi-Squared Data | In a thought-provoking paper, Efron (2011) investigated the merit and
limitation of an empirical Bayes method to correct selection bias based on
Tweedie's formula first reported by \cite{Robbins:1956}. The exceptional virtue
of Tweedie's formula for the normal distribution lies in its representation of
selection bias as a simple function of the derivative of log marginal
likelihood. Since the marginal likelihood and its derivative can be estimated
from the data directly without invoking prior information, bias correction can
be carried out conveniently. We propose a Bayesian hierarchical model for
chi-squared data such that the resulting Tweedie's formula has the same virtue
as that of the normal distribution. Because the family of noncentral
chi-squared distributions, the common alternative distributions for chi-squared
tests, does not constitute an exponential family, our results cannot be
obtained by extending existing results. Furthermore, the corresponding
Tweedie's formula manifests new phenomena quite different from those of the
normal distribution and suggests new ways of analyzing chi-squared data.
| stat |
On the Equivalence between Online and Private Learnability beyond Binary
Classification | Alon et al. [2019] and Bun et al. [2020] recently showed that online
learnability and private PAC learnability are equivalent in binary
classification. We investigate whether this equivalence extends to multi-class
classification and regression. First, we show that private learnability implies
online learnability in both settings. Our extension involves studying a novel
variant of the Littlestone dimension that depends on a tolerance parameter and
on an appropriate generalization of the concept of threshold functions beyond
binary classification. Second, we show that while online learnability continues
to imply private learnability in multi-class classification, current proof
techniques encounter significant hurdles in the regression setting. While the
equivalence for regression remains open, we provide non-trivial sufficient
conditions for an online learnable class to also be privately learnable.
| stat |
Neural ODEs as the Deep Limit of ResNets with constant weights | In this paper we prove that, in the deep limit, the stochastic gradient
descent on a ResNet type deep neural network, where each layer shares the same
weight matrix, converges to the stochastic gradient descent for a Neural ODE
and that the corresponding value/loss functions converge. Our result gives, in
the context of minimization by stochastic gradient descent, a theoretical
foundation for considering Neural ODEs as the deep limit of ResNets. Our proof
is based on certain decay estimates for associated Fokker-Planck equations.
| stat |
Improving estimation of the volume under the ROC surface when data are
missing not at random | In this paper, we propose a mean score equation-based approach to estimate
the the volume under the receiving operating characteristic (ROC) surface (VUS)
of a diagnostic test, under nonignorable (NI) verification bias. The proposed
approach involves a parametric regression model for the verification process,
which accommodates for possible NI missingness in the disease status of sample
subjects, and may use instrumental variables, which help avoid possible
identifiability problems. In order to solve the mean score equation derived by
the chosen verification model, we preliminarily need to estimate the parameters
of a model for the disease process, but its specification is required only for
verified subjects under study. Then, by using the estimated verification and
disease probabilities, we obtain four verification bias-corrected VUS
estimators, which are alternative to those recently proposed by To Duc et al.
(2019), based on a full likelihood approach. Consistency and asymptotic
normality of the new estimators are established. Simulation experiments are
conducted to evaluate their finite sample performances, and an application to a
dataset from a research on epithelial ovarian cancer is presented.
| stat |
Coding for Crowdsourced Classification with XOR Queries | This paper models the crowdsourced labeling/classification problem as a
sparsely encoded source coding problem, where each query answer, regarded as a
code bit, is the XOR of a small number of labels, as source information bits.
In this paper we leverage the connections between this problem and well-studied
codes with sparse representations for the channel coding problem to provide
querying schemes with almost optimal number of queries, each of which involving
only a constant number of labels. We also extend this scenario to the case
where some workers can be unresponsive. For this case, we propose querying
schemes where each query involves only log n items, where n is the total number
of items to be labeled. Furthermore, we consider classification of two
correlated labeling systems and provide two-stage querying schemes with almost
optimal number of queries each involving a constant number of labels.
| stat |
Synthesizing geocodes to facilitate access to detailed geographical
information in large scale administrative data | We investigate whether generating synthetic data can be a viable strategy for
providing access to detailed geocoding information for external researchers,
without compromising the confidentiality of the units included in the database.
Our work was motivated by a recent project at the Institute for Employment
Research (IAB) in Germany that linked exact geocodes to the Integrated
Employment Biographies (IEB), a large administrative database containing
several million records. We evaluate the performance of three synthesizers
regarding the trade-off between preserving analytical validity and limiting
disclosure risks: One synthesizer employs Dirichlet Process mixtures of
products of multinomials (DPMPM), while the other two use different versions of
Classification and Regression Trees (CART). In terms of preserving analytical
validity, our proposed synthesis strategy for geocodes based on categorical
CART models outperforms the other two. If the risks of the synthetic data
generated by the categorical CART synthesizer are deemed too high, we
demonstrate that synthesizing additional variables is the preferred strategy to
address the risk-utility trade-off in practice, compared to limiting the size
of the regression trees or relying on the strategy of providing geographical
information only on an aggregated level. We also propose strategies for making
the synthesizers scalable for large files, present analytical validity measures
and disclosure risk measures for the generated data, and provide general
recommendations for statistical agencies considering the synthetic data
approach for disseminating detailed geographical information.
| stat |
Glucodensities: a new representation of glucose profiles using
distributional data analysis | Biosensor data has the potential ability to improve disease control and
detection. However, the analysis of these data under free-living conditions is
not feasible with current statistical techniques. To address this challenge, we
introduce a new functional representation of biosensor data, termed the
glucodensity, together with a data analysis framework based on distances
between them. The new data analysis procedure is illustrated through an
application in diabetes with continuous-time glucose monitoring (CGM) data. In
this domain, we show marked improvement with respect to state of the art
analysis methods. In particular, our findings demonstrate that i) the
glucodensity possesses an extraordinary clinical sensitivity to capture the
typical biomarkers used in the standard clinical practice in diabetes, ii)
previous biomarkers cannot accurately predict glucodensity, so that the latter
is a richer source of information, and iii) the glucodensity is a natural
generalization of the time in range metric, this being the gold standard in the
handling of CGM data. Furthermore, the new method overcomes many of the
drawbacks of time in range metrics, and provides deeper insight into assessing
glucose metabolism.
| stat |
A Fast and Scalable Implementation Method for Competing Risks Data with
the R Package fastcmprsk | Advancements in medical informatics tools and high-throughput biological
experimentation make large-scale biomedical data routinely accessible to
researchers. Competing risks data are typical in biomedical studies where
individuals are at risk to more than one cause (type of event) which can
preclude the others from happening. The Fine-Gray model is a popular and
well-appreciated model for competing risks data and is currently implemented in
a number of statistical software packages. However, current implementations are
not computationally scalable for large-scale competing risks data. We have
developed an R package, fastcmprsk, that uses a novel forward-backward scan
algorithm to significantly reduce the computational complexity for parameter
estimation by exploiting the structure of the subject-specific risk sets.
Numerical studies compare the speed and scalability of our implementation to
current methods for unpenalized and penalized Fine-Gray regression and show
impressive gains in computational efficiency.
| stat |
A matrix-free approach to geostatistical filtering | In this paper, we present a novel approach to geostatistical filtering which
tackles two challenges encountered when applying this method to complex spatial
datasets: modeling the non-stationarity of the data while still being able to
work with large datasets. The approach is based on a finite element
approximation of Gaussian random fields expressed as an expansion of the
eigenfunctions of a Laplace--Beltrami operator defined to account for local
anisotropies. The numerical approximation of the resulting random fields using
a finite element approach is then leveraged to solve the scalability issue
through a matrix-free approach. Finally, two cases of application of this
approach, on simulated and real seismic data are presented.
| stat |
A Practical Model-based Segmentation Approach for Accurate Activation
Detection in Single-Subject functional Magnetic Resonance Imaging Studies | Functional Magnetic Resonance Imaging (fMRI) maps cerebral activation in
response to stimuli but this activation is often difficult to detect,
especially in low-signal contexts and single-subject studies. Accurate
activation detection can be guided by the fact that very few voxels are, in
reality, truly activated and that activated voxels are spatially localized, but
it is challenging to incorporate both these facts. We provide a computationally
feasible and methodologically sound model-based approach, implemented in the R
package MixfMRI, that bounds the a priori expected proportion of activated
voxels while also incorporating spatial context. Results on simulation
experiments for different levels of activation detection difficulty are
uniformly encouraging. The value of the methodology in low-signal and
single-subject fMRI studies is illustrated on a sports imagination experiment.
Concurrently, we also extend the potential use of fMRI as a clinical tool to,
for example, detect awareness and improve treatment in individual patients in
persistent vegetative state, such as traumatic brain injury survivors.
| stat |
Spatial trend analysis of gridded temperature data at varying spatial
scales | Classical assessments of trends in gridded temperature data perform
independent evaluations across the grid, thus, ignoring spatial correlations in
the trend estimates. In particular, this affects assessments of trend
significance as evaluation of the collective significance of individual tests
is commonly neglected. In this article we build a space-time hierarchical
Bayesian model for temperature anomalies where the trend coefficient is modeled
by a latent Gaussian random field. This enables us to calculate simultaneous
credible regions for joint significance assessments. In a case study, we assess
summer season trends in 65 years of gridded temperature data over Europe. We
find that while spatial smoothing generally results in larger regions where the
null hypothesis of no trend is rejected, this is not the case for all
sub-regions.
| stat |
GRATIS: GeneRAting TIme Series with diverse and controllable
characteristics | The explosion of time series data in recent years has brought a flourish of
new time series analysis methods, for forecasting, clustering, classification
and other tasks. The evaluation of these new methods requires either collecting
or simulating a diverse set of time series benchmarking data to enable reliable
comparisons against alternative approaches. We propose GeneRAting TIme Series
with diverse and controllable characteristics, named GRATIS, with the use of
mixture autoregressive (MAR) models. We simulate sets of time series using MAR
models and investigate the diversity and coverage of the generated time series
in a time series feature space. By tuning the parameters of the MAR models,
GRATIS is also able to efficiently generate new time series with controllable
features. In general, as a costless surrogate to the traditional data
collection approach, GRATIS can be used as an evaluation tool for tasks such as
time series forecasting and classification. We illustrate the usefulness of our
time series generation process through a time series forecasting application.
| stat |
Design and analysis considerations for a sequentially randomized HIV
prevention trial | TechStep is a randomized trial of a mobile health interventions targeted
towards transgender adolescents. The interventions include a short message
system, a mobile-optimized web application, and electronic counseling. The
primary outcomes are self-reported sexual risk behaviors and uptake of HIV
preventing medication. In order that we may evaluate the efficacy of several
different combinations of interventions, the trial has a sequentially
randomized design. We use a causal framework to formalize the estimands of the
primary and key secondary analyses of the TechStep trial data. Targeted minimum
loss-based estimators of these quantities are described and studied in
simulation.
| stat |
Meta-learning with Stochastic Linear Bandits | We investigate meta-learning procedures in the setting of stochastic linear
bandits tasks. The goal is to select a learning algorithm which works well on
average over a class of bandits tasks, that are sampled from a
task-distribution. Inspired by recent work on learning-to-learn linear
regression, we consider a class of bandit algorithms that implement a
regularized version of the well-known OFUL algorithm, where the regularization
is a square euclidean distance to a bias vector. We first study the benefit of
the biased OFUL algorithm in terms of regret minimization. We then propose two
strategies to estimate the bias within the learning-to-learn setting. We show
both theoretically and experimentally, that when the number of tasks grows and
the variance of the task-distribution is small, our strategies have a
significant advantage over learning the tasks in isolation.
| stat |
High-Dimensional Poisson DAG Model Learning Using $\ell_1$-Regularized
Regression | In this paper, we develop a new approach to learning high-dimensional Poisson
directed acyclic graphical (DAG) models from only observational data without
strong assumptions such as faithfulness and strong sparsity. A key component of
our method is to decouple the ordering estimation or parent search where the
problems can be efficiently addressed using $\ell_1$-regularized regression and
the mean-variance relationship. We show that sample size $n = \Omega( d^{2}
\log^{9} p)$ is sufficient for our polynomial time Mean-variance Ratio Scoring
(MRS) algorithm to recover the true directed graph, where $p$ is the number of
nodes and $d$ is the maximum indegree. We verify through simulations that our
algorithm is statistically consistent in the high-dimensional $p>n$ setting,
and performs well compared to state-of-the-art ODS, GES, and MMHC algorithms.
We also demonstrate through multivariate real count data that our MRS algorithm
is well-suited to estimating DAG models for multivariate count data in
comparison to other methods used for discrete data.
| stat |
'In-Between' Uncertainty in Bayesian Neural Networks | We describe a limitation in the expressiveness of the predictive uncertainty
estimate given by mean-field variational inference (MFVI), a popular
approximate inference method for Bayesian neural networks. In particular, MFVI
fails to give calibrated uncertainty estimates in between separated regions of
observations. This can lead to catastrophically overconfident predictions when
testing on out-of-distribution data. Avoiding such overconfidence is critical
for active learning, Bayesian optimisation and out-of-distribution robustness.
We instead find that a classical technique, the linearised Laplace
approximation, can handle 'in-between' uncertainty much better for small
network architectures.
| stat |
Variable selection with multiply-imputed datasets: choosing between
stacked and grouped methods | Penalized regression methods, such as lasso and elastic net, are used in many
biomedical applications when simultaneous regression coefficient estimation and
variable selection is desired. However, missing data complicates the
implementation of these methods, particularly when missingness is handled using
multiple imputation. Applying a variable selection algorithm on each imputed
dataset will likely lead to different sets of selected predictors, making it
difficult to ascertain a final active set without resorting to ad hoc
combination rules. In this paper we consider a general class of penalized
objective functions which, by construction, force selection of the same
variables across multiply-imputed datasets. By pooling objective functions
across imputations, optimization is then performed jointly over all imputed
datasets rather than separately for each dataset. We consider two objective
function formulations that exist in the literature, which we will refer to as
"stacked" and "grouped" objective functions. Building on existing work, we (a)
derive and implement efficient cyclic coordinate descent and
majorization-minimization optimization algorithms for both continuous and
binary outcome data, (b) incorporate adaptive shrinkage penalties, (c) compare
these methods through simulation, and (d) develop an R package miselect for
easy implementation. Simulations demonstrate that the "stacked" objective
function approaches tend to be more computationally efficient and have better
estimation and selection properties. We apply these methods to data from the
University of Michigan ALS Patients Repository (UMAPR) which aims to identify
the association between persistent organic pollutants and ALS risk.
| stat |
Assessing Impact of Unobserved Confounders with Sensitivity Index
Probabilities through Pseudo-Experiments | Unobserved confounders are a long-standing issue in causal inference using
propensity score methods. This study proposed nonparametric indices to quantify
the impact of unobserved confounders through pseudo-experiments with an
application to real-world data. The study finding suggests that the proposed
indices can reflect the true impact of confounders. It is hoped that this study
will lead to further discussion on this important issue and help move the
science of causal inference forward.
| stat |
High-dimensional outlier detection using random projections | There exist multiple methods to detect outliers in multivariate data in the
literature, but most of them require to estimate the covariance matrix. The
higher the dimension, the more complex the estimation of the matrix becoming
impossible in high dimensions. In order to avoid estimating this matrix, we
propose a novel random projections-based procedure to detect outliers in
Gaussian multivariate data. It consists in projecting the data in several
one-dimensional subspaces where an appropriate univariate outlier detection
method, similar to Tukey's method but with a threshold depending on the initial
dimension and the sample size, is applied. The required number of projections
is determined using sequential analysis. Simulated and real datasets illustrate
the performance of the proposed method.
| stat |
Fitting High-Dimensional Interaction Models with Error Control | There is a renewed interest in polynomial regression in the form of
identifying influential interactions between features. In many settings, this
takes place in a high-dimensional model, making the number of interactions
unwieldy or computationally infeasible. Furthermore, it is difficult to analyze
such spaces directly as they are often highly correlated. Standard feature
selection issues remain such as how to determine a final model which
generalizes well. This paper solves these problems with a sequential algorithm
called Revisiting Alpha-Investing (RAI). RAI is motivated by the principle of
marginality and searches the feature-space of higher-order interactions by
greedily building upon lower-order terms. RAI controls a notion of false
rejections and comes with a performance guarantee relative to the best-subset
model. This ensures that signal is identified while providing a valid stopping
criterion to prevent over-selection. We apply RAI in a novel setting over a
family of regressions in order to select gene-specific interaction models for
differential expression profiling.
| stat |
Time-Series Imputation with Wasserstein Interpolation for Optimal
Look-Ahead-Bias and Variance Tradeoff | Missing time-series data is a prevalent practical problem. Imputation methods
in time-series data often are applied to the full panel data with the purpose
of training a model for a downstream out-of-sample task. For example, in
finance, imputation of missing returns may be applied prior to training a
portfolio optimization model. Unfortunately, this practice may result in a
look-ahead-bias in the future performance on the downstream task. There is an
inherent trade-off between the look-ahead-bias of using the full data set for
imputation and the larger variance in the imputation from using only the
training data. By connecting layers of information revealed in time, we propose
a Bayesian posterior consensus distribution which optimally controls the
variance and look-ahead-bias trade-off in the imputation. We demonstrate the
benefit of our methodology both in synthetic and real financial data.
| stat |
Estimating prediction error for complex samples | With a growing interest in using non-representative samples to train
prediction models for numerous outcomes it is necessary to account for the
sampling design that gives rise to the data in order to assess the generalized
predictive utility of a proposed prediction rule. After learning a prediction
rule based on a non-uniform sample, it is of interest to estimate the rule's
error rate when applied to unobserved members of the population. Efron (1986)
proposed a general class of covariance penalty inflated prediction error
estimators that assume the available training data are representative of the
target population for which the prediction rule is to be applied. We extend
Efron's estimator to the complex sample context by incorporating
Horvitz-Thompson sampling weights and show that it is consistent for the true
generalization error rate when applied to the underlying superpopulation. The
resulting Horvitz-Thompson-Efron (HTE) estimator is equivalent to dAIC, a
recent extension of AIC to survey sampling data, but is more widely applicable.
The proposed methodology is assessed with simulations and is applied to models
predicting renal function obtained from the large-scale NHANES survey.
| stat |
Sequential Monte Carlo Methods in the nimble R Package | nimble is an R package for constructing algorithms and conducting inference
on hierarchical models. The nimble package provides a unique combination of
flexible model specification and the ability to program model-generic
algorithms. Specifically, the package allows users to code models in the BUGS
language, and it allows users to write algorithms that can be applied to any
appropriate model. In this paper, we introduce nimble's capabilities for
state-space model analysis using sequential Monte Carlo (SMC) techniques. We
first provide an overview of state-space models and commonly-used SMC
algorithms. We then describe how to build a state-space model and conduct
inference using existing SMC algorithms within nimble. SMC algorithms within
nimble currently include the bootstrap filter, auxiliary particle filter,
ensemble Kalman filter, IF2 method of iterated filtering, and a particle MCMC
sampler. These algorithms can be run in R or compiled into C++ for more
efficient execution. Examples of applying SMC algorithms to linear
autoregressive models and a stochastic volatility model are provided. Finally,
we give an overview of how model-generic algorithms are coded within nimble by
providing code for a simple SMC algorithm. This illustrates how users can
easily extend nimble's SMC methods in high-level code.
| stat |
A likelihood analysis of quantile-matching transformations | Quantile matching is a strictly monotone transformation that sends the
observed response values $\{y_1, . . . , y_n\}$ to the quantiles of a given
target distribution. A likelihood based criterion is developed for comparing
one target distribution with another in a linear-model setting.
| stat |
Non-exponentially weighted aggregation: regret bounds for unbounded loss
functions | We tackle the problem of online optimization with a general, possibly
unbounded, loss function. It is well known that when the loss is bounded, the
exponentially weighted aggregation strategy (EWA) leads to a regret in
$\sqrt{T}$ after $T$ steps. In this paper, we study a generalized aggregation
strategy, where the weights do no longer depend exponentially on the losses.
Our strategy is defined as the minimizer of the expected losses plus a
regularizer. When this regularizer is the Kullback-Leibler divergence, we
obtain EWA as a special case, but using alternative divergences lead to regret
bounds for unbounded losses, at the cost of a worst regret bound in some cases.
| stat |
Online control of the familywise error rate | Biological research often involves testing a growing number of null
hypotheses as new data is accumulated over time. We study the problem of online
control of the familywise error rate (FWER), that is testing an apriori
unbounded sequence of hypotheses (p-values) one by one over time without
knowing the future, such that with high probability there are no false
discoveries in the entire sequence. This paper unifies algorithmic concepts
developed for offline (single batch) FWER control and online false discovery
rate control to develop novel online FWER control methods. Though many offline
FWER methods (e.g. Bonferroni, fallback procedures and Sidak's method) can
trivially be extended to the online setting, our main contribution is the
design of new, powerful, adaptive online algorithms that control the FWER when
the p-values are independent or locally dependent in time. Our experiments
demonstrate substantial gains in power, that are also formally proved in a
Gaussian sequence model. Multiple testing, FWER control, online setting.
| stat |
TenIPS: Inverse Propensity Sampling for Tensor Completion | Tensors are widely used to represent multiway arrays of data. The recovery of
missing entries in a tensor has been extensively studied, generally under the
assumption that entries are missing completely at random (MCAR). However, in
most practical settings, observations are missing not at random (MNAR): the
probability that a given entry is observed (also called the propensity) may
depend on other entries in the tensor or even on the value of the missing
entry. In this paper, we study the problem of completing a partially observed
tensor with MNAR observations, without prior information about the
propensities. To complete the tensor, we assume that both the original tensor
and the tensor of propensities have low multilinear rank. The algorithm first
estimates the propensities using a convex relaxation and then predicts missing
values using a higher-order SVD approach, reweighting the observed tensor by
the inverse propensities. We provide finite-sample error bounds on the
resulting complete tensor. Numerical experiments demonstrate the effectiveness
of our approach.
| stat |
On a family of discrete log-symmetric distributions | The use of continuous probability distributions has been widespread in
problems with purely discrete nature. In general, such distributions are not
appropriate in this scenario. In this paper, we introduce a class of discrete
and asymmetric distributions based on the family of continuous log-symmetric
distributions. Some properties are discussed as well as estimation by the
maximum likelihood method. A Monte Carlo simulation study is carried out to
evaluate the performance of the estimators, and censored and uncensored data
sets are used to illustrate the proposed methodology.
| stat |
Correspondence Analysis between the Location and the Leading Causes of
Death in the United States | Correspondence Analysis analyzes two-way or multi-way tables withe each row
and column becoming a point ion a multidimensional graphical map called biplot.
It can be used to extract essential dimensions allowing simplification of the
data matrix. This study aims to measure the association between the location
and the leading causes of death in the United States of America and to
determine the location where a particular disease is highly associated. The
research data consists of two variables with 510 data points. Results show that
there is a significant association between the location ad leading cause of
death in the United States, and 61% of the variance in the model are explained
by the first two dimensions.
| stat |
On general notions of depth for regression | Depth notions in location have fascinated tremendous attention in the
literature. In fact data depth and its applications remain one of the most
active research topics in statistics in the last two decades. Most favored
notions of depth in location include Tukey (1975) halfspace depth (HD), Liu
(1990) simplicial depth, and projection depth (Stahel (1981) and Donoho (1982),
Liu (1992), Zuo and Serfling (2000) (ZS00) and Zuo (2003)), among others.
Depth notions in regression have also been proposed, sporadically
nevertheless. Regression depth (RD) of Rousseeuw and Hubert (1999) (RH99) is
the most famous one which is a direct extension of Tukey HD to regression.
Others include Carrizosa (1996) and the ones induced from Marrona and Yohai
(1993) (MY93) proposed in this article. Is there any relationship between
Carrizosa depth and RD of RH99? Do these depth notions possess desirable
properties? What are the desirable properties? Can existing notions really
serve as depth functions in regression? These questions remain open.
Revealing the equivalence between Carrizosa depth and RD of RH99; expanding
location depth evaluating criteria in ZS00 for regression depth notions;
examining the existing regression notions with respect to the gauges; and
proposing the regression counterpart of the eminent projection depth in
location are the four major objectives of the article.
| stat |
Necessary and sufficient conditions for causal feature selection in time
series with latent common causes | We study the identification of direct and indirect causes on time series and
provide conditions in the presence of latent variables, which we prove to be
necessary and sufficient under some graph constraints. Our theoretical results
and estimation algorithms require two conditional independence tests for each
observed candidate time series to determine whether or not it is a cause of an
observed target time series. We provide experimental results in simulations, as
well as real data. Our results show that our method leads to very low false
positives and relatively low false negative rates, outperforming the widely
used Granger causality.
| stat |
RGISTools: Downloading, Customizing, and Processing Time Series of
Remote Sensing Data in R | There is a large number of data archives and web services offering free
access to multispectral satellite imagery. Images from multiple sources are
increasingly combined to improve the spatio-temporal coverage of measurements
while achieving more accurate results. Archives and web services differ in
their protocols, formats, and data standards, which are barriers to combine
datasets. Here, we present RGISTools, an R package to create time-series of
multispectral satellite images from multiple platforms in a harmonized and
standardized way. We first provide an overview of the package functionalities,
namely downloading, customizing, and processing multispectral satellite imagery
for a region and time period of interest as well as a recent statistical method
for gap-filling and smoothing series of images, called interpolation of the
mean anomalies. We further show the capabilities of the package through a case
study that combines Landsat-8 and Sentinel-2 satellite optical imagery to
estimate the level of a water reservoir in Northern Spain. We expect RGISTools
to foster research on data fusion and spatio-temporal modelling using satellite
images from multiple programs.
| stat |
Comments on "Deep Neural Networks with Random Gaussian Weights: A
Universal Classification Strategy?" | In a recently published paper [1], it is shown that deep neural networks
(DNNs) with random Gaussian weights preserve the metric structure of the data,
with the property that the distance shrinks more when the angle between the two
data points is smaller. We agree that the random projection setup considered in
[1] preserves distances with a high probability. But as far as we are
concerned, the relation between the angle of the data points and the output
distances is quite the opposite, i.e., smaller angles result in a weaker
distance shrinkage. This leads us to conclude that Theorem 3 and Figure 5 in
[1] are not accurate. Hence the usage of random Gaussian weights in DNNs cannot
provide an ability of universal classification or treating in-class and
out-of-class data separately. Consequently, the behavior of networks consisting
of random Gaussian weights only is not useful to explain how DNNs achieve
state-of-art results in a large variety of problems.
| stat |
Olympus: a benchmarking framework for noisy optimization and experiment
planning | Research challenges encountered across science, engineering, and economics
can frequently be formulated as optimization tasks. In chemistry and materials
science, recent growth in laboratory digitization and automation has sparked
interest in optimization-guided autonomous discovery and closed-loop
experimentation. Experiment planning strategies based on off-the-shelf
optimization algorithms can be employed in fully autonomous research platforms
to achieve desired experimentation goals with the minimum number of trials.
However, the experiment planning strategy that is most suitable to a scientific
discovery task is a priori unknown while rigorous comparisons of different
strategies are highly time and resource demanding. As optimization algorithms
are typically benchmarked on low-dimensional synthetic functions, it is unclear
how their performance would translate to noisy, higher-dimensional experimental
tasks encountered in chemistry and materials science. We introduce Olympus, a
software package that provides a consistent and easy-to-use framework for
benchmarking optimization algorithms against realistic experiments emulated via
probabilistic deep-learning models. Olympus includes a collection of
experimentally derived benchmark sets from chemistry and materials science and
a suite of experiment planning strategies that can be easily accessed via a
user-friendly python interface. Furthermore, Olympus facilitates the
integration, testing, and sharing of custom algorithms and user-defined
datasets. In brief, Olympus mitigates the barriers associated with benchmarking
optimization algorithms on realistic experimental scenarios, promoting data
sharing and the creation of a standard framework for evaluating the performance
of experiment planning strategies
| stat |
Marginal Densities, Factor Graph Duality, and High-Temperature Series
Expansions | We prove that the marginal densities of a global probability mass function in
a primal normal factor graph and the corresponding marginal densities in the
dual normal factor graph are related via local mappings. The mapping depends on
the Fourier transform of the local factors of the models. Details of the
mapping, including its fixed points, are derived for the Ising model, and then
extended to the Potts model. By employing the mapping, we can transform
simultaneously all the estimated marginal densities from one domain to the
other, which is advantageous if estimating the marginals can be carried out
more efficiently in the dual domain. An example of particular significance is
the ferromagnetic Ising model in a positive external field, for which there is
a rapidly mixing Markov chain (called the subgraphs-world process) to generate
configurations in the dual normal factor graph of the model. Our numerical
experiments illustrate that the proposed procedure can provide more accurate
estimates of marginal densities in various settings.
| stat |
Deep generative models in DataSHIELD | The best way to calculate statistics from medical data is to use the data of
individual patients. In some settings, this data is difficult to obtain due to
privacy restrictions. In Germany, for example, it is not possible to pool
routine data from different hospitals for research purposes without the consent
of the patients. The DataSHIELD software provides an infrastructure and a set
of statistical methods for joint analyses of distributed data. The contained
algorithms are reformulated to work with aggregated data from the participating
sites instead of the individual data. If a desired algorithm is not implemented
in DataSHIELD or cannot be reformulated in such a way, using artificial data is
an alternative. We present a methodology together with a software
implementation that builds on DataSHIELD to create artificial data that
preserve complex patterns from distributed individual patient data. Such data
sets of artificial patients, which are not linked to real patients, can then be
used for joint analyses. We use deep Boltzmann machines (DBMs) as generative
models for capturing the distribution of data. For the implementation, we
employ the package "BoltzmannMachines" from the Julia programming language and
wrap it for use with DataSHIELD, which is based on R. As an exemplary
application, we conduct a distributed analysis with DBMs on a synthetic data
set, which simulates genetic variant data. Patterns from the original data can
be recovered in the artificial data using hierarchical clustering of the
virtual patients, demonstrating the feasibility of the approach. Our
implementation adds to DataSHIELD the ability to generate artificial data that
can be used for various analyses, e. g. for pattern recognition with deep
learning. This also demonstrates more generally how DataSHIELD can be flexibly
extended with advanced algorithms from languages other than R.
| stat |
Classification Accuracy and Parameter Estimation in Multilevel Contexts:
A Study of Conditional Nonparametric Multilevel Latent Class Analysis | The current research has two aims. First, to demonstrate the utility
conditional nonparametric multilevel latent class analysis (NP-MLCA) for
multi-site program evaluation using an empirical dataset. Second, to
investigate how classification accuracy and parameter estimation of a
conditional NP-MLCA are affected by six study factors: the quality of latent
class indicators, the number of latent class indicators, level-1 covariate
effects, cross-level covariate effects, the number of level-2 units, and the
size of level-2 units. A total of 96 conditions was examined using a simulation
study. The resulting classification accuracy rates, the power and type-I error
of cross-level covariate effects and contextual effects suggest that the
nonparametric multilevel latent class model can be applied broadly in
multilevel contexts.
| stat |
Partial least squares for sparsely observed curves with measurement
errors | Functional partial least squares (FPLS) is commonly used for fitting
scalar-on-function regression models. For the sake of accuracy, FPLS demands
that each realization of the functional predictor is recorded as densely as
possible over the entire time span; however, this condition is sometimes
violated in, e.g., longitudinal studies and missing data research. Targeting
this point, we adapt FPLS to scenarios in which the number of measurements per
subject is small and bounded from above. The resulting proposal is abbreviated
as PLEASS. Under certain regularity conditions, we establish the consistency of
estimators and give confidence intervals for scalar responses. Simulation
studies and real-data applications illustrate the competitive accuracy of
PLEASS
| stat |
Nonparametric additive factor models | We study an additive factor model in which the common components are related
to the latent factors through unknown smooth functions. The model is new to the
literature and generalizes nonlinear parametric factor models. Identification
conditions are provided and a general nonparametric estimation procedure is
specified. Convergence of the estimator is obtained when both time and
cross-sectional size increase at appropriate rates. The finite sample
performance is then illustrated in extensive numerical experiments. Finally,
the relevance and usefulness of the method is shown in an application to a
nonlinear CAPM on S&P500 data.
| stat |
Statistical Inference for Online Decision-Making: In a Contextual Bandit
Setting | Online decision-making problem requires us to make a sequence of decisions
based on incremental information. Common solutions often need to learn a reward
model of different actions given the contextual information and then maximize
the long-term reward. It is meaningful to know if the posited model is
reasonable and how the model performs in the asymptotic sense. We study this
problem under the setup of the contextual bandit framework with a linear reward
model. The $\varepsilon$-greedy policy is adopted to address the classic
exploration-and-exploitation dilemma. Using the martingale central limit
theorem, we show that the online ordinary least squares estimator of model
parameters is asymptotically normal. When the linear model is misspecified, we
propose the online weighted least squares estimator using the inverse
propensity score weighting and also establish its asymptotic normality. Based
on the properties of the parameter estimators, we further show that the
in-sample inverse propensity weighted value estimator is asymptotically normal.
We illustrate our results using simulations and an application to a news
article recommendation dataset from Yahoo!.
| stat |
A Simple Correction Procedure for High-Dimensional Generalized Linear
Models with Measurement Error | We consider high-dimensional generalized linear models when the covariates
are contaminated by measurement error. Estimates from errors-in-variables
regression models are well-known to be biased in traditional low-dimensional
settings if the error is unincorporated. Such models have recently become of
interest when regularizing penalties are added to the estimation procedure.
Unfortunately, correcting for the mismeasurements can add undue computational
difficulties onto the optimization, which a new tool set for practitioners to
successfully use the models. We investigate a general procedure that utilizes
the recently proposed Imputation-Regularized Optimization algorithm for
high-dimensional errors-in-variables models, which we implement for continuous,
binary, and count response type. Crucially, our method allows for off-the-shelf
linear regression methods to be employed in the presence of contaminated
covariates. We apply our correction to gene microarray data, and illustrate
that it results in a great reduction in the number of false positives whilst
still retaining most true positives.
| stat |
A new integrated likelihood for estimating population size in dependent
dual-record system | Efficient estimation of population size from dependent dual-record system
(DRS) remains a statistical challenge in capture-recapture type experiment.
Owing to the nonidentifiability of the suitable Time-Behavioral Response
Variation model (denoted as $M_{tb}$) under DRS, few methods are developed in
Bayesian paradigm based on informative priors. Our contribution in this article
is in developing integrated likelihood function from model $M_{tb}$ based on a
novel approach developed by Severini (2007, Biometrika). Suitable weight
function on nuisance parameter is derived under the assumption of availability
of knowledge on the direction of behavioral dependency. Such pseudo-likelihood
function is constructed so that the resulting estimator possess some desirable
properties including invariance and negligible prior (or weight) sensitiveness.
Extensive simulations explore the better performance of our proposed method in
most of the situations than the existing Bayesian methods. Moreover, being a
non-Bayesian estimator, it simply avoids heavy computational effort and time.
Finally, illustration based on two real life data sets on epidemiology and
economic census are presented.
| stat |
Provably Training Neural Network Classifiers under Fairness Constraints | Training a classifier under fairness constraints has gotten increasing
attention in the machine learning community thanks to moral, legal, and
business reasons. However, several recent works addressing algorithmic fairness
have only focused on simple models such as logistic regression or support
vector machines due to non-convex and non-differentiable fairness criteria
across protected groups, such as race or gender. Neural networks, the most
widely used models for classification nowadays, are precluded and lack
theoretical guarantees. This paper aims to fill this missing but crucial part
of the literature of algorithmic fairness for neural networks. In particular,
we show that overparametrized neural networks could meet the fairness
constraints. The key ingredient of building a fair neural network classifier is
establishing no-regret analysis for neural networks in the overparameterization
regime, which may be of independent interest in the online learning of neural
networks and related applications.
| stat |
BIVA: A Very Deep Hierarchy of Latent Variables for Generative Modeling | With the introduction of the variational autoencoder (VAE), probabilistic
latent variable models have received renewed attention as powerful generative
models. However, their performance in terms of test likelihood and quality of
generated samples has been surpassed by autoregressive models without
stochastic units. Furthermore, flow-based models have recently been shown to be
an attractive alternative that scales well to high-dimensional data. In this
paper we close the performance gap by constructing VAE models that can
effectively utilize a deep hierarchy of stochastic variables and model complex
covariance structures. We introduce the Bidirectional-Inference Variational
Autoencoder (BIVA), characterized by a skip-connected generative model and an
inference network formed by a bidirectional stochastic inference path. We show
that BIVA reaches state-of-the-art test likelihoods, generates sharp and
coherent natural images, and uses the hierarchy of latent variables to capture
different aspects of the data distribution. We observe that BIVA, in contrast
to recent results, can be used for anomaly detection. We attribute this to the
hierarchy of latent variables which is able to extract high-level semantic
features. Finally, we extend BIVA to semi-supervised classification tasks and
show that it performs comparably to state-of-the-art results by generative
adversarial networks.
| stat |
Generalised Network Autoregressive Processes and the GNAR package | This article introduces the GNAR package, which fits, predicts, and simulates
from a powerful new class of generalised network autoregressive processes. Such
processes consist of a multivariate time series along with a real, or inferred,
network that provides information about inter-variable relationships. The GNAR
model relates values of a time series for a given variable and time to earlier
values of the same variable and of neighbouring variables, with inclusion
controlled by the network structure. The GNAR package is designed to fit this
new model, while working with standard ts objects and the igraph package for
ease of use.
| stat |
Towards an AI assistant for human grid operators | Power systems are becoming more complex to operate in the digital age. As a
result, real-time decision-making is getting more challenging as the human
operator has to deal with more information, more uncertainty, more applications
and more coordination. While supervision has been primarily used to help them
make decisions over the last decades, it cannot reasonably scale up anymore.
There is a great need for rethinking the human-machine interface under more
unified and interactive frameworks. Taking advantage of the latest developments
in Human-machine Interactions and Artificial intelligence, we share the vision
of a new assistant framework relying on an hypervision interface and greater
bidirectional interactions. We review the known principles of decision-making
that drives the assistant design and supporting assistance functions we
present. We finally share some guidelines to make progress towards the
development of such an assistant.
| stat |
LassoNet: A Neural Network with Feature Sparsity | Much work has been done recently to make neural networks more interpretable,
and one obvious approach is to arrange for the network to use only a subset of
the available features. In linear models, Lasso (or $\ell_1$-regularized)
regression assigns zero weights to the most irrelevant or redundant features,
and is widely used in data science. However the Lasso only applies to linear
models. Here we introduce LassoNet, a neural network framework with global
feature selection. Our approach enforces a hierarchy: specifically a feature
can participate in a hidden unit only if its linear representative is active.
Unlike other approaches to feature selection for neural nets, our method uses a
modified objective function with constraints, and so integrates feature
selection with the parameter learning directly. As a result, it delivers an
entire regularization path of solutions with a range of feature sparsity. On
systematic experiments, LassoNet significantly outperforms state-of-the-art
methods for feature selection and regression. The LassoNet method uses
projected proximal gradient descent, and generalizes directly to deep networks.
It can be implemented by adding just a few lines of code to a standard neural
network.
| stat |
Meta Learning as Bayes Risk Minimization | Meta-Learning is a family of methods that use a set of interrelated tasks to
learn a model that can quickly learn a new query task from a possibly small
contextual dataset. In this study, we use a probabilistic framework to
formalize what it means for two tasks to be related and reframe the
meta-learning problem into the problem of Bayesian risk minimization (BRM). In
our formulation, the BRM optimal solution is given by the predictive
distribution computed from the posterior distribution of the task-specific
latent variable conditioned on the contextual dataset, and this justifies the
philosophy of Neural Process. However, the posterior distribution in Neural
Process violates the way the posterior distribution changes with the contextual
dataset. To address this problem, we present a novel Gaussian approximation for
the posterior distribution that generalizes the posterior of the linear
Gaussian model. Unlike that of the Neural Process, our approximation of the
posterior distributions converges to the maximum likelihood estimate with the
same rate as the true posterior distribution. We also demonstrate the
competitiveness of our approach on benchmark datasets.
| stat |
Robust Principal Component Analysis Based On Maximum Correntropy Power
Iterations | Principal component analysis (PCA) is recognised as a quintessential data
analysis technique when it comes to describing linear relationships between the
features of a dataset. However, the well-known sensitivity of PCA to
non-Gaussian samples and/or outliers often makes it unreliable in practice. To
this end, a robust formulation of PCA is derived based on the maximum
correntropy criterion (MCC) so as to maximise the expected likelihood of
Gaussian distributed reconstruction errors. In this way, the proposed solution
reduces to a generalised power iteration, whereby: (i) robust estimates of the
principal components are obtained even in the presence of outliers; (ii) the
number of principal components need not be specified in advance; and (iii) the
entire set of principal components can be obtained, unlike existing approaches.
The advantages of the proposed maximum correntropy power iteration (MCPI) are
demonstrated through an intuitive numerical example.
| stat |
Fast, Optimal, and Targeted Predictions using Parametrized Decision
Analysis | Prediction is critical for decision-making under uncertainty and lends
validity to statistical inference. With targeted prediction, the goal is to
optimize predictions for specific decision tasks of interest, which we
represent via functionals. Although classical decision analysis extracts
predictions from a Bayesian model, these predictions are often difficult to
interpret and slow to compute. Instead, we design a class of parametrized
actions for Bayesian decision analysis that produce optimal, scalable, and
simple targeted predictions. For a wide variety of action parametrizations and
loss functions--including linear actions with sparsity constraints for targeted
variable selection--we derive a convenient representation of the optimal
targeted prediction that yields efficient and interpretable solutions.
Customized out-of-sample predictive metrics are developed to evaluate and
compare among targeted predictors. Through careful use of the posterior
predictive distribution, we introduce a procedure that identifies a set of
near-optimal, or acceptable targeted predictors, which provide unique insights
into the features and level of complexity needed for accurate targeted
prediction. Simulations demonstrate excellent prediction, estimation, and
variable selection capabilities. Targeted predictions are constructed for
physical activity data from the National Health and Nutrition Examination
Survey (NHANES) to better predict and understand the characteristics of
intraday physical activity.
| stat |
Bivariate Gamma Mixture of Experts Models for Joint Insurance Claims
Modeling | In general insurance, risks from different categories are often modeled
independently and their sum is regarded as the total risk the insurer takes on
in exchange for a premium. The dependence from multiple risks is generally
neglected even when correlation could exist, for example a single car accident
may result in claims from multiple risk categories. It is desirable to take the
covariance of different categories into consideration in modeling in order to
better predict future claims and hence allow greater accuracy in ratemaking. In
this work multivariate severity models are investigated using mixture of
experts models with bivariate gamma distributions, where the dependence
structure is modeled directly using a GLM framework, and covariates can be
placed in both gating and expert networks. Furthermore, parsimonious
parameterisations are considered, which leads to a family of bivariate gamma
mixture of experts models. It can be viewed as a model-based clustering
approach that clusters policyholders into sub-groups with different
dependencies, and the parameters of the mixture models are dependent on the
covariates. Clustering is shown to be important in separating the data into
sub-groupings where strong dependence is often present, even if the overall
data set exhibits only weak dependence. In doing so, the correlation within
different components features prominently in the model. It is shown that, by
applying to both simulated data and a real-world Irish GI insurer data set,
claim predictions can be improved.
| stat |
Mixture composite regression models with multi-type feature selection | The aim of this paper is to present a mixture composite regression model for
claim severity modelling. Claim severity modelling poses several challenges
such as multimodality, heavy-tailedness and systematic effects in data. We
tackle this modelling problem by studying a mixture composite regression model
for simultaneous modeling of attritional and large claims, and for considering
systematic effects in both the mixture components as well as the mixing
probabilities. For model fitting, we present a group-fused regularization
approach that allows us for selecting the explanatory variables which
significantly impact the mixing probabilities and the different mixture
components, respectively. We develop an asymptotic theory for this regularized
estimation approach, and fitting is performed using a novel Generalized
Expectation-Maximization algorithm. We exemplify our approach on real motor
insurance data set.
| stat |
Fighting Contextual Bandits with Stochastic Smoothing | We introduce a new stochastic smoothing perspective to study adversarial
contextual bandit problems. We propose a general algorithm template that
represents random perturbation based algorithms and identify several
perturbation distributions that lead to strong regret bounds. Using the idea of
smoothness, we provide an $O(\sqrt{T})$ zero-order bound for the vanilla
algorithm and an $O(L^{*2/3}_{T})$ first-order bound for the clipped version.
These bounds hold when the algorithms use with a variety of distributions that
have a bounded hazard rate. Our algorithm template includes EXP4 as a special
case corresponding to the Gumbel perturbation. Our regret bounds match existing
results for EXP4 without relying on the specific properties of the algorithm.
| stat |
Causal mechanism of extreme river discharges in the upper Danube basin
network | Extreme hydrological events in the Danube river basin may severely impact
human populations, aquatic organisms, and economic activity. One often
characterizes the joint structure of the extreme events using the theory of
multivariate and spatial extremes and its asymptotically justified models.
There is interest however in cascading extreme events and whether one event
causes another. In this paper, we argue that an improved understanding of the
mechanism underlying severe events is achieved by combining extreme value
modelling and causal discovery. We construct a causal inference method relying
on the notion of the Kolmogorov complexity of extreme conditional quantiles.
Tail quantities are derived using multivariate extreme value models and
causal-induced asymmetries in the data are explored through the minimum
description length principle. Our CausEV, for Causality for Extreme Values,
approach uncovers causal relations between summer extreme river discharges in
the upper Danube basin and finds significant causal links between the Danube
and its Alpine tributary Lech.
| stat |
Sets of Half-Average Nulls Generate Risk-Limiting Audits: SHANGRLA | Risk-limiting audits (RLAs) for many social choice functions can be reduced
to testing sets of null hypotheses of the form "the average of this list is not
greater than 1/2" for a collection of finite lists of nonnegative numbers. Such
social choice functions include majority, super-majority, plurality,
multi-winner plurality, Instant Runoff Voting (IRV), Borda count, approval
voting, and STAR-Voting, among others. The audit stops without a full hand
count iff all the null hypotheses are rejected. The nulls can be tested in many
ways. Ballot-polling is particularly simple; two new ballot-polling
risk-measuring functions for sampling without replacement are given.
Ballot-level comparison audits transform each null into an equivalent assertion
that the mean of re-scaled tabulation errors is not greater than 1/2. In turn,
that null can then be tested using the same statistical methods used for ballot
polling---but applied to different finite lists of nonnegative numbers.
SHANGRLA comparison audits are more efficient than previous comparison audits
for two reasons: (i) for most social choice functions, the conditions tested
are both necessary and sufficient for the reported outcome to be correct, while
previous methods tested conditions that were sufficient but not necessary, and
(ii) the tests avoid a conservative approximation. The SHANGRLA abstraction
simplifies stratified audits, including audits that combine ballot polling with
ballot-level comparisons, producing sharper audits than the "SUITE" approach.
SHANGRLA works with the "phantoms to evil zombies" strategy to treat missing
ballot cards and missing or redacted cast vote records. That also facilitates
sampling from "ballot-style manifests," which can dramatically improve
efficiency when the audited contests do not appear on every ballot card.
Open-source software implementing SHANGRLA ballot-level comparison audits is
available.
| stat |
AdaptSPEC-X: Covariate Dependent Spectral Modeling of Multiple
Nonstationary Time Series | We present a method for the joint analysis of a panel of possibly
nonstationary time series. The approach is Bayesian and uses a
covariate-dependent infinite mixture model to incorporate multiple time series,
with mixture components parameterized by a time varying mean and log spectrum.
The mixture components are based on AdaptSPEC, a nonparametric model which
adaptively divides the time series into an unknown number of segments and
estimates the local log spectra by smoothing splines. We extend AdaptSPEC to
handle missing values, a common feature of time series which can cause
difficulties for nonparametric spectral methods. A second extension is to allow
for a time varying mean. Covariates, assumed to be time-independent, are
incorporated via the mixture weights using the logistic stick breaking process.
The model can estimate time varying means and spectra at observed and
unobserved covariate values, allowing for predictive inference. Estimation is
performed by Markov chain Monte Carlo (MCMC) methods, combining data
augmentation, reversible jump, and Riemann manifold Hamiltonian Monte Carlo
techniques. We evaluate the methodology using simulated data, and describe
applications to Australian rainfall data and measles incidence in the US.
Software implementing the method proposed in this paper is available in the R
package BayesSpec.
| stat |
Understanding Implicit Regularization in Over-Parameterized Nonlinear
Statistical Model | We study the implicit regularization phenomenon induced by simple
optimization algorithms in over-parameterized nonlinear statistical models.
Specifically, we study both vector and matrix single index models where the
link function is nonlinear and unknown, the signal parameter is either a sparse
vector or a low-rank symmetric matrix, and the response variable can be
heavy-tailed. To gain a better understanding of the role played by implicit
regularization in the nonlinear models without excess technicality, we assume
that the distribution of the covariates is known a priori. For both the vector
and matrix settings, we construct an over-parameterized least-squares loss
function by employing the score function transform and a robust truncation step
designed specifically for heavy-tailed data. We propose to estimate the true
parameter by applying regularization-free gradient descent to the loss
function. When the initialization is close to the origin and the stepsize is
sufficiently small, we prove that the obtained solution achieves minimax
optimal statistical rates of convergence in both the vector and matrix cases.
In particular, for the vector single index model with Gaussian covariates, our
proposed estimator is shown to further enjoy the oracle statistical rate. Our
results capture the implicit regularization phenomenon in over-parameterized
nonlinear and noisy statistical models with possibly heavy-tailed data.
| stat |
Confidence Sets and Hypothesis Testing in a Likelihood-Free Inference
Setting | Parameter estimation, statistical tests and confidence sets are the
cornerstones of classical statistics that allow scientists to make inferences
about the underlying process that generated the observed data. A key question
is whether one can still construct hypothesis tests and confidence sets with
proper coverage and high power in a so-called likelihood-free inference (LFI)
setting; that is, a setting where the likelihood is not explicitly known but
one can forward-simulate observable data according to a stochastic model. In
this paper, we present $\texttt{ACORE}$ (Approximate Computation via Odds Ratio
Estimation), a frequentist approach to LFI that first formulates the classical
likelihood ratio test (LRT) as a parametrized classification problem, and then
uses the equivalence of tests and confidence sets to build confidence regions
for parameters of interest. We also present a goodness-of-fit procedure for
checking whether the constructed tests and confidence regions are valid.
$\texttt{ACORE}$ is based on the key observation that the LRT statistic, the
rejection probability of the test, and the coverage of the confidence set are
conditional distribution functions which often vary smoothly as a function of
the parameters of interest. Hence, instead of relying solely on samples
simulated at fixed parameter settings (as is the convention in standard Monte
Carlo solutions), one can leverage machine learning tools and data simulated in
the neighborhood of a parameter to improve estimates of quantities of interest.
We demonstrate the efficacy of $\texttt{ACORE}$ with both theoretical and
empirical results. Our implementation is available on Github.
| stat |
Addressing cluster-constant covariates in mixed effects models via
likelihood-based boosting techniques | Boosting techniques from the field of statistical learning have grown to be a
popular tool for estimating and selecting predictor effects in various
regression models and can roughly be separated in two general approaches,
namely gradient boosting and likelihood-based boosting. An extensive framework
has been proposed in order to fit generalised mixed models based on boosting,
however for the case of cluster-constant covariates likelihood-based boosting
approaches tend to mischoose variables in the selection step leading to wrong
estimates. We propose an improved boosting algorithm for linear mixed models
where the random effects are properly weighted, disentangled from the fixed
effects updating scheme and corrected for correlations with cluster-constant
covariates in order to improve quality of estimates and in addition reduce the
computational effort. The method outperforms current state-of-the-art
approaches from boosting and maximum likelihood inference which is shown via
simulations and various data examples.
| stat |
Counterfactual Representation Learning with Balancing Weights | A key to causal inference with observational data is achieving balance in
predictive features associated with each treatment type. Recent literature has
explored representation learning to achieve this goal. In this work, we discuss
the pitfalls of these strategies - such as a steep trade-off between achieving
balance and predictive power - and present a remedy via the integration of
balancing weights in causal learning. Specifically, we theoretically link
balance to the quality of propensity estimation, emphasize the importance of
identifying a proper target population, and elaborate on the complementary
roles of feature balancing and weight adjustments. Using these concepts, we
then develop an algorithm for flexible, scalable and accurate estimation of
causal effects. Finally, we show how the learned weighted representations may
serve to facilitate alternative causal learning procedures with appealing
statistical features. We conduct an extensive set of experiments on both
synthetic examples and standard benchmarks, and report encouraging results
relative to state-of-the-art baselines.
| stat |
View selection in multi-view stacking: Choosing the meta-learner | Multi-view stacking is a framework for combining information from different
views (i.e. different feature sets) describing the same set of objects. In this
framework, a base-learner algorithm is trained on each view separately, and
their predictions are then combined by a meta-learner algorithm. In a previous
study, stacked penalized logistic regression, a special case of multi-view
stacking, has been shown to be useful in identifying which views are most
important for prediction. In this article we expand this research by
considering seven different algorithms to use as the meta-learner, and
evaluating their view selection and classification performance in simulations
and two applications on real gene-expression data sets. Our results suggest
that if both view selection and classification accuracy are important to the
research at hand, then the nonnegative lasso, nonnegative adaptive lasso and
nonnegative elastic net are suitable meta-learners. Exactly which among these
three is to be preferred depends on the research context. The remaining four
meta-learners, namely nonnegative ridge regression, nonnegative forward
selection, stability selection and the interpolating predictor, show little
advantages in order to be preferred over the other three.
| stat |
A Permutation Test for Assessing the Presence of Individual Differences
in Treatment Effects | One size fits all approaches to medicine have become a thing of the past as
the understanding of individual differences grows. The paper introduces a test
for the presence of heterogeneity in treatment effects in a clinical trial.
Heterogeneity is assessed on the basis of the predicted individual treatment
effects (PITE) framework and a permutation test is utilized to establish if
significant heterogeneity is present. We first use the novel test to show that
heterogeneity in the effects of interventions exists in the Amyotrophic Lateral
Sclerosis Clinical Trials. We then show, using two different predictive models
(linear regression model and Random Forests) that the test has adequate type I
error control. Next, we use the ALS data as the basis for simulations to
demonstrate the ability of the permutation test to find heterogeneity in
treatment effects as a function of both effect size and sample size. We find
that the proposed test has good power to detected heterogeneity in treatment
effects when the heterogeneity was due primarily to a single predictor, or when
it was spread across the predictors. The predictive model, on the other hand is
of secondary importance to detect heterogeneity. The non-parametric property of
the permutation test can be applied with any predictive method and requires no
additional assumptions to obtain PITEs.
| stat |
A Causal Bootstrap | The bootstrap, introduced by Efron (1982), has become a very popular method
for estimating variances and constructing confidence intervals. A key insight
is that one can approximate the properties of estimators by using the empirical
distribution function of the sample as an approximation for the true
distribution function. This approach views the uncertainty in the estimator as
coming exclusively from sampling uncertainty. We argue that for causal
estimands the uncertainty arises entirely, or partially, from a different
source, corresponding to the stochastic nature of the treatment received. We
develop a bootstrap procedure that accounts for this uncertainty, and compare
its properties to that of the classical bootstrap.
| stat |
Weak-Instrument Robust Tests in Two-Sample Summary-Data Mendelian
Randomization | Mendelian randomization (MR) is a popular method in genetic epidemiology to
estimate the effect of an exposure on an outcome using genetic variants as
instrumental variables (IV), with two-sample summary-data MR being the most
popular. Unfortunately, instruments in MR studies are often weakly associated
with the exposure, which can bias effect estimates and inflate Type I errors.
In this work, we propose test statistics that are robust under weak instrument
asymptotics by extending the Anderson-Rubin, Kleibergen, and conditional
likelihood ratio tests in econometrics to two-sample summary-data MR. We also
use the proposed Anderson-Rubin test to develop a point estimator and to detect
invalid instruments. We conclude with a simulation and an empirical study and
show that the proposed tests control size and have better power than existing
methods.
| stat |
Smooth Online Parameter Estimation for time varying VAR models with
application to rat's LFP data | Multivariate time series data appear often as realizations of non-stationary
processes where the covariance matrix or spectral matrix smoothly evolve over
time. Most of the current approaches estimate the time-varying spectral
properties only retrospectively - that is, after the entire data has been
observed. Retrospective estimation is a major limitation in many adaptive
control applications where it is important to estimate these properties and
detect changes in the system as they happen in real-time. One major obstacle in
online estimation is the computational cost due to the high-dimensionality of
the parameters. Existing methods such as the Kalman filter or local least
squares are feasible. However, they are not always suitable because they
provide noisy estimates and can become prohibitively costly as the dimension of
the time series increases. In our brain signal application, it is critical to
develop a robust method that can estimate, in real-time, the properties of the
underlying stochastic process, in particular, the spectral brain connectivity
measures. For these reasons we propose a new smooth online parameter estimation
approach (SOPE) that has the ability to control for the smoothness of the
estimates with a reasonable computational complexity. Consequently, the models
are fit in real-time even for high dimensional time series. We demonstrate that
our proposed SOPE approach is as good as the Kalman filter in terms of
mean-squared error for small dimensions. However, unlike the Kalman filter, the
SOPE has lower computational cost and hence scalable for higher dimensions.
Finally, we apply the SOPE method to a rat's local field potential data during
a hippocampus-dependent sequence-memory task. As demonstrated in the video, the
proposed SOPE method is able to capture the dynamics of the connectivity as the
rat performs the sequence of non-spatial working memory tasks.
| stat |
Causal network learning with non-invertible functional relationships | Discovery of causal relationships from observational data is an important
problem in many areas. Several recent results have established the
identifiability of causal DAGs with non-Gaussian and/or nonlinear structural
equation models (SEMs). In this paper, we focus on nonlinear SEMs defined by
non-invertible functions, which exist in many data domains, and propose a novel
test for non-invertible bivariate causal models. We further develop a method to
incorporate this test in structure learning of DAGs that contain both linear
and nonlinear causal relations. By extensive numerical comparisons, we show
that our algorithms outperform existing DAG learning methods in identifying
causal graphical structures. We illustrate the practical application of our
method in learning causal networks for combinatorial binding of transcription
factors from ChIP-Seq data.
| stat |
Generalised Joint Regression for Count Data with a Focus on Modelling
Football Matches | We propose a versatile joint regression framework for count responses. The
method is implemented in the R add-on package GJRM and allows for modelling
linear and non-linear dependence through the use of several copulae. Moreover,
the parameters of the marginal distributions of the count responses and of the
copula can be specified as flexible functions of covariates. Motivated by a
football application, we also discuss an extension which forces the regression
coefficients of the marginal (linear) predictors to be equal via a suitable
penalisation. Model fitting is based on a trust region algorithm which
estimates simultaneously all the parameters of the joint models. We investigate
the proposal's empirical performance in two simulation studies, the first one
designed for arbitrary count data, the other one reflecting football-specific
settings. Finally, the method is applied to FIFA World Cup data, showing its
competitiveness to the standard approach with regard to predictive performance.
| stat |
On the Convergence of Memory-Based Distributed SGD | Distributed stochastic gradient descent~(DSGD) has been widely used for
optimizing large-scale machine learning models, including both convex and
non-convex models. With the rapid growth of model size, huge communication cost
has been the bottleneck of traditional DSGD. Recently, many communication
compression methods have been proposed. Memory-based distributed stochastic
gradient descent~(M-DSGD) is one of the efficient methods since each worker
communicates a sparse vector in each iteration so that the communication cost
is small. Recent works propose the convergence rate of M-DSGD when it adopts
vanilla SGD. However, there is still a lack of convergence theory for M-DSGD
when it adopts momentum SGD. In this paper, we propose a universal convergence
analysis for M-DSGD by introducing \emph{transformation equation}. The
transformation equation describes the relation between traditional DSGD and
M-DSGD so that we can transform M-DSGD to its corresponding DSGD. Hence we get
the convergence rate of M-DSGD with momentum for both convex and non-convex
problems. Furthermore, we combine M-DSGD and stagewise learning that the
learning rate of M-DSGD in each stage is a constant and is decreased by stage,
instead of iteration. Using the transformation equation, we propose the
convergence rate of stagewise M-DSGD which bridges the gap between theory and
practice.
| stat |
Posterior Distribution for the Number of Clusters in Dirichlet Process
Mixture Models | Dirichlet process mixture models (DPMM) play a central role in Bayesian
nonparametrics, with applications throughout statistics and machine learning.
DPMMs are generally used in clustering problems where the number of clusters is
not known in advance, and the posterior distribution is treated as providing
inference for this number. Recently, however, it has been shown that the DPMM
is inconsistent in inferring the true number of components in certain cases.
This is an asymptotic result, and it would be desirable to understand whether
it holds with finite samples, and to more fully understand the full posterior.
In this work, we provide a rigorous study for the posterior distribution of the
number of clusters in DPMM under different prior distributions on the
parameters and constraints on the distributions of the data. We provide novel
lower bounds on the ratios of probabilities between $s+1$ clusters and $s$
clusters when the prior distributions on parameters are chosen to be Gaussian
or uniform distributions.
| stat |
Probabilistic Trajectory Segmentation by Means of Hierarchical Dirichlet
Process Switching Linear Dynamical Systems | Using movement primitive libraries is an effective means to enable robots to
solve more complex tasks. In order to build these movement libraries, current
algorithms require a prior segmentation of the demonstration trajectories. A
promising approach is to model the trajectory as being generated by a set of
Switching Linear Dynamical Systems and inferring a meaningful segmentation by
inspecting the transition points characterized by the switching dynamics. With
respect to the learning, a nonparametric Bayesian approach is employed
utilizing a Gibbs sampler.
| stat |
Modelling the spatial extent and severity of extreme European windstorms | Windstorms are a primary natural hazard affecting Europe that are commonly
linked to substantial property and infrastructural damage and are responsible
for the largest spatially aggregated financial losses. Such extreme winds are
typically generated by extratropical cyclone systems originating in the North
Atlantic and passing over Europe. Previous statistical studies tend to model
extreme winds at a given set of sites, corresponding to inference in a Eulerian
framework. Such inference cannot incorporate knowledge of the life cycle and
progression of extratropical cyclones across the region and is forced to make
restrictive assumptions about the extremal dependence structure. We take an
entirely different approach which overcomes these limitations by working in a
Lagrangian framework. Specifically, we model the development of windstorms over
time, preserving the physical characteristics linking the windstorm and the
cyclone track, the path of local vorticity maxima, and make a key finding that
the spatial extent of extratropical windstorms becomes more localised as its
magnitude increases irrespective of the location of the storm track. Our model
allows simulation of synthetic windstorm events to derive the joint
distributional features over any set of sites giving physically consistent
extrapolations to rarer events. From such simulations improved estimates of
this hazard can be achieved both in terms of intensity and area affected.
| stat |
Objective Bayesian approach to the Jeffreys-Lindley paradox | We consider the Jeffreys-Lindley paradox from an objective Bayesian
perspective by attempting to find priors representing complete indifference to
sample size in the problem. This means that we ensure that the prior for the
unknown mean and the prior predictive for the $t$-statistic are independent of
the sample size. If successful, this would lead to Bayesian model comparison
that was independent of sample size and ameliorate the paradox. Unfortunately,
it leads to an improper scale-invariant prior for the unknown mean. We show,
however, that a truncated scale-invariant prior delays the dependence on sample
size, which could be practically significant. Lastly, we shed light on the
paradox by relating it to the fact that the scale-invariant prior is improper.
| stat |
Deep Learning for Classification Tasks on Geospatial Vector Polygons | In this paper, we evaluate the accuracy of deep learning approaches on
geospatial vector geometry classification tasks. The purpose of this evaluation
is to investigate the ability of deep learning models to learn from geometry
coordinates directly. Previous machine learning research applied to geospatial
polygon data did not use geometries directly, but derived properties thereof.
These are produced by way of extracting geometry properties such as Fourier
descriptors. Instead, our introduced deep neural net architectures are able to
learn on sequences of coordinates mapped directly from polygons. In three
classification tasks we show that the deep learning architectures are
competitive with common learning algorithms that require extracted features.
| stat |
Deviation bound for non-causal machine learning | Concentration inequalities are widely used for analyzing machine learning
algorithms. However, current concentration inequalities cannot be applied to
some of the most popular deep neural networks, notably in natural language
processing. This is mostly due to the non-causal nature of such involved data,
in the sense that each data point depends on other neighbor data points. In
this paper, a framework for modeling non-causal random fields is provided and a
Hoeffding-type concentration inequality is obtained for this framework. The
proof of this result relies on a local approximation of the non-causal random
field by a function of a finite number of i.i.d. random variables.
| stat |
Budget-limited distribution learning in multifidelity problems | Multifidelity methods are widely used for statistical estimation of
quantities of interest (QoIs) in uncertainty quantification using simulation
codes of differing costs and accuracies. Many methods approximate
numerical-valued statistics that represent only limited information of the
QoIs. In this paper, we introduce a semi-parametric approach that aims to
effectively describe the distribution of a scalar-valued QoI in the
multifidelity setup. Under a linear model hypothesis, we propose an
exploration-exploitation strategy to reconstruct the full distribution of a
scalar-valued QoI using samples from a subset of low-fidelity regressors. We
derive an informative asymptotic bound for the mean 1-Wasserstein distance
between the estimator and the true distribution, and use it to adaptively
allocate computational budget for parametric estimation and non-parametric
reconstruction. Assuming the linear model is correct, we prove that such a
procedure is consistent, and converges to the optimal policy (and hence optimal
computational budget allocation) under an upper bound criterion as the budget
goes to infinity. A major advantage of our approach compared to several other
multifidelity methods is that it is automatic, and its implementation does not
require a hierarchical model setup, cross-model information, or \textit{a
priori} known model statistics. Numerical experiments are provided in the end
to support our theoretical analysis.
| stat |
Causal inference with limited resources: proportionally-representative
interventions | Investigators often evaluate treatment effects by considering settings in
which all individuals are assigned a treatment of interest, assuming that an
unlimited number of treatment units are available. However, many real-life
treatments are of limited supply and cannot be provided to all individuals in
the population. For example, patients on the liver transplant waiting list
cannot be assigned a liver transplant immediately at the time they reach
highest priority because a suitable organ is not likely to be immediately
available. In these cases, investigators may still be interested in the effects
of treatment strategies in which a finite number of organs are available at a
given time, that is, treatment regimes that satisfy resource constraints. Here,
we describe an estimand that can be used to define causal effects of treatment
strategies that satisfy resource constraints: proportionally-representative
interventions for limited resources. We derive a simple class of inverse
probability weighted estimators, and apply one such estimator to evaluate the
effect of restricting or expanding utilization of "increased risk" liver organs
to treat patients with end-stage liver disease. Our method is designed to
evaluate policy-relevant interventions in the setting of finite treatment
resources.
| stat |
Population-level Task-evoked Functional Connectivity | Functional magnetic resonance imaging (fMRI) is a non-invasive, in-vivo
imaging technique essential for measuring brain activity. Functional
connectivity is used to study associations between brain regions either at rest
or while study participants perform tasks. This paper proposes a rigorous
definition of task-evoked functional connectivity at the population level
(ptFC). Importantly, our proposed ptFC is interpretable in the context of
task-fMRI studies. Two algorithms for estimating ptFC are provided. We present
the performance of the proposed algorithms compared to existing functional
connectivity estimation approaches using simulations. Lastly, we apply the
proposed framework to estimate functional connectivity in a motor-task study
from the Human Connectome Project.
| stat |
When effects cannot be estimated: redefining estimands to understand the
effects of naloxone access laws | Background: All states in the US have enacted at least some naloxone access
laws (NALs) in an effort to reduce opioid overdose lethality. Previous
evaluations found NALs increased naloxone dispensing but showed mixed results
in terms of opioid overdose mortality. One reason for mixed results could be
failure to address violations of the positivity assumption caused by the
co-occurrence of NAL enactment with enactment of related laws, ultimately
resulting in bias, increased variance, and low statistical power.
Methods: We reformulated the research question to alleviate some challenges
related to law co-occurrence. Because NAL enactment was closely correlated with
Good Samaritan Law (GSL) enactment, we bundled NAL with GSL, and estimated the
hypothetical associations of enacting NAL/GSL up to 2 years earlier (an amount
supported by the observed data) on naloxone dispensation and opioid overdose
mortality.
Results: We estimated that such a shift in NAL/GSL duration would have been
associated with increased naloxone dispensations (0.28 dispensations/100,000
people (95% CI: 0.18-0.38) in 2013 among early NAL/GSL enactors; 47.58 (95% CI:
28.40-66.76) in 2018 among late enactors). We estimated that such a shift would
have been associated with increased opioid overdose mortality (1.88
deaths/100,000 people (95% CI: 1.03-2.72) in 2013; 2.06 (95% CI: 0.92-3.21) in
2018).
Conclusions: Consistent with prior research, increased duration of NAL/GSL
enactment was associated with increased naloxone dispensing. Contrary to
expectation, we did not find a protective association with opioid overdose
morality, though residual bias due to unobserved confounding and interference
likely remain.
| stat |
Modeling Shared Responses in Neuroimaging Studies through MultiView ICA | Group studies involving large cohorts of subjects are important to draw
general conclusions about brain functional organization. However, the
aggregation of data coming from multiple subjects is challenging, since it
requires accounting for large variability in anatomy, functional topography and
stimulus response across individuals. Data modeling is especially hard for
ecologically relevant conditions such as movie watching, where the experimental
setup does not imply well-defined cognitive operations.
We propose a novel MultiView Independent Component Analysis (ICA) model for
group studies, where data from each subject are modeled as a linear combination
of shared independent sources plus noise. Contrary to most group-ICA
procedures, the likelihood of the model is available in closed form. We develop
an alternate quasi-Newton method for maximizing the likelihood, which is robust
and converges quickly. We demonstrate the usefulness of our approach first on
fMRI data, where our model demonstrates improved sensitivity in identifying
common sources among subjects. Moreover, the sources recovered by our model
exhibit lower between-session variability than other methods.On
magnetoencephalography (MEG) data, our method yields more accurate source
localization on phantom data. Applied on 200 subjects from the Cam-CAN dataset
it reveals a clear sequence of evoked activity in sensor and source space.
The code is freely available at https://github.com/hugorichard/multiviewica.
| stat |
The Limits to Learning a Diffusion Model | This paper provides the first sample complexity lower bounds for the
estimation of simple diffusion models, including the Bass model (used in
modeling consumer adoption) and the SIR model (used in modeling epidemics). We
show that one cannot hope to learn such models until quite late in the
diffusion. Specifically, we show that the time required to collect a number of
observations that exceeds our sample complexity lower bounds is large. For Bass
models with low innovation rates, our results imply that one cannot hope to
predict the eventual number of adopting customers until one is at least
two-thirds of the way to the time at which the rate of new adopters is at its
peak. In a similar vein, our results imply that in the case of an SIR model,
one cannot hope to predict the eventual number of infections until one is
approximately two-thirds of the way to the time at which the infection rate has
peaked. These limits are borne out in both product adoption data (Amazon), as
well as epidemic data (COVID-19).
| stat |
Bayesian Mixed Effects Models for Zero-inflated Compositions in
Microbiome Data Analysis | Detecting associations between microbial compositions and sample
characteristics is one of the most important tasks in microbiome studies. Most
of the existing methods apply univariate models to single microbial species
separately, with adjustments for multiple hypothesis testing. We propose a
Bayesian analysis for a generalized mixed effects linear model tailored to this
application. The marginal prior on each microbial composition is a Dirichlet
Process, and dependence across compositions is induced through a linear
combination of individual covariates, such as disease biomarkers or the
subject's age, and latent factors. The latent factors capture residual
variability and their dimensionality is learned from the data in a fully
Bayesian procedure. The proposed model is tested in data analyses and
simulation studies with zero-inflated compositions. In these settings, within
each sample, a large proportion of counts per microbial species are equal to
zero. In our Bayesian model a priori the probability of compositions with
absent microbial species is strictly positive. We propose an efficient
algorithm to sample from the posterior and visualizations of model parameters
which reveal associations between covariates and microbial compositions. We
evaluate the proposed method in simulation studies, and then analyze a
microbiome dataset for infants with type 1 diabetes which contains a large
proportion of zeros in the sample-specific microbial compositions.
| stat |
Parametric Gaussian Process Regressors | The combination of inducing point methods with stochastic variational
inference has enabled approximate Gaussian Process (GP) inference on large
datasets. Unfortunately, the resulting predictive distributions often exhibit
substantially underestimated uncertainties. Notably, in the regression case the
predictive variance is typically dominated by observation noise, yielding
uncertainty estimates that make little use of the input-dependent function
uncertainty that makes GP priors attractive. In this work we propose two simple
methods for scalable GP regression that address this issue and thus yield
substantially improved predictive uncertainties. The first applies variational
inference to FITC (Fully Independent Training Conditional; Snelson
et.~al.~2006). The second bypasses posterior approximations and instead
directly targets the posterior predictive distribution. In an extensive
empirical comparison with a number of alternative methods for scalable GP
regression, we find that the resulting predictive distributions exhibit
significantly better calibrated uncertainties and higher log likelihoods--often
by as much as half a nat per datapoint.
| stat |
Scalable Exact Inference in Multi-Output Gaussian Processes | Multi-output Gaussian processes (MOGPs) leverage the flexibility and
interpretability of GPs while capturing structure across outputs, which is
desirable, for example, in spatio-temporal modelling. The key problem with
MOGPs is their computational scaling $O(n^3 p^3)$, which is cubic in the number
of both inputs $n$ (e.g., time points or locations) and outputs $p$. For this
reason, a popular class of MOGPs assumes that the data live around a
low-dimensional linear subspace, reducing the complexity to $O(n^3 m^3)$.
However, this cost is still cubic in the dimensionality of the subspace $m$,
which is still prohibitively expensive for many applications. We propose the
use of a sufficient statistic of the data to accelerate inference and learning
in MOGPs with orthogonal bases. The method achieves linear scaling in $m$ in
practice, allowing these models to scale to large $m$ without sacrificing
significant expressivity or requiring approximation. This advance opens up a
wide range of real-world tasks and can be combined with existing GP
approximations in a plug-and-play way. We demonstrate the efficacy of the
method on various synthetic and real-world data sets.
| stat |
A Hybrid Framework for Topology Identification of Distribution Grid with
Renewables Integration | Topology identification (TI) is a key task for state estimation (SE) in
distribution grids, especially the one with high-penetration renewables. The
uncertainties, initiated by the time-series behavior of renewables, will almost
certainly lead to bad TI results without a proper treatment. These
uncertainties are analytically intractable under conventional framework--they
are usually jointly spatial-temporal dependent, and hence cannot be simply
treated as white noise. For this purpose, a hybrid framework is suggested in
this paper to handle these uncertainties in a systematic and theoretical way;
in particular, big data analytics are studied to harness the jointly
spatial-temporal statistical properties of those uncertainties. With some prior
knowledge, a model bank is built first to store the countable typical models of
network configurations; therefore, the difference between the SE outputs of
each bank model and our observation is capable of being defined as a matrix
variate--the so-called random matrix. In order to gain insight into the random
matrix, a well-designed metric space is needed. Auto-regression (AR) model,
factor analysis (FA), and random matrix theory (RMT) are tied together for the
metric space design, followed by jointly temporal-spatial analysis of those
matrices which is conducted in a high-dimensional (vector) space. Under the
proposed framework, some big data analytics and theoretical results are
obtained to improve the TI performance. Our framework is validated using IEEE
standard distribution network with some field data in practice.
| stat |
Data-Space Inversion Using a Recurrent Autoencoder for Time-Series
Parameterization | Data-space inversion (DSI) and related procedures represent a family of
methods applicable for data assimilation in subsurface flow settings. These
methods differ from model-based techniques in that they provide only posterior
predictions for quantities (time series) of interest, not posterior models with
calibrated parameters. DSI methods require a large number of flow simulations
to first be performed on prior geological realizations. Given observed data,
posterior predictions can then be generated directly. DSI operates in a
Bayesian setting and provides posterior samples of the data vector. In this
work we develop and evaluate a new approach for data parameterization in DSI.
Parameterization reduces the number of variables to determine in the inversion,
and it maintains the physical character of the data variables. The new
parameterization uses a recurrent autoencoder (RAE) for dimension reduction,
and a long-short-term memory (LSTM) network to represent flow-rate time series.
The RAE-based parameterization is combined with an ensemble smoother with
multiple data assimilation (ESMDA) for posterior generation. Results are
presented for two- and three-phase flow in a 2D channelized system and a 3D
multi-Gaussian model. The RAE procedure, along with existing DSI treatments,
are assessed through comparison to reference rejection sampling (RS) results.
The new DSI methodology is shown to consistently outperform existing
approaches, in terms of statistical agreement with RS results. The method is
also shown to accurately capture derived quantities, which are computed from
variables considered directly in DSI. This requires correlation and covariance
between variables to be properly captured, and accuracy in these relationships
is demonstrated. The RAE-based parameterization developed here is clearly
useful in DSI, and it may also find application in other subsurface flow
problems.
| stat |
Multiple competition-based FDR control for peptide detection | Competition-based FDR control has been commonly used for over a decade in the
computational mass spectrometry community (Elias and Gygi, 2007). Recently, the
approach has gained significant popularity in other fields after Barber and
Candes (2015) laid its theoretical foundation in a more general setting that
included the feature selection problem. In both cases, the competition is based
on a head-to-head comparison between an observed score and a corresponding
decoy / knockoff. Keich and Noble (2017b) recently demonstrated some advantages
of using multiple rather than a single decoy when addressing the problem of
assigning peptide sequences to observed mass spectra. In this work, we consider
a related problem -- detecting peptides based on a collection of mass spectra
-- and we develop a new framework for competition-based FDR control using
multiple null scores. Within this framework, we offer several methods, all of
which are based on a novel procedure that rigorously controls the FDR in the
finite sample setting. Using real data to study the peptide detection problem
we show that, relative to existing single-decoy methods, our approach can
increase the number of discovered peptides by up to 50% at small FDR
thresholds.
| stat |
All Sparse PCA Models Are Wrong, But Some Are Useful. Part I:
Computation of Scores, Residuals and Explained Variance | Sparse Principal Component Analysis (sPCA) is a popular matrix factorization
approach based on Principal Component Analysis (PCA) that combines variance
maximization and sparsity with the ultimate goal of improving data
interpretation. When moving from PCA to sPCA, there are a number of
implications that the practitioner needs to be aware of. A relevant one is that
scores and loadings in sPCA may not be orthogonal. For this reason, the
traditional way of computing scores, residuals and variance explained that is
used in the classical PCA cannot directly be applied to sPCA models. This also
affects how sPCA components should be visualized. In this paper we illustrate
this problem both theoretically and numerically using simulations for several
state-of-the-art sPCA algorithms, and provide proper computation of the
different elements mentioned. We show that sPCA approaches present disparate
and limited performance when modeling noise-free, sparse data. In a follow-up
paper, we discuss the theoretical properties that lead to this problem.
| stat |
Compressed Sensing with Deep Image Prior and Learned Regularization | We propose a novel method for compressed sensing recovery using untrained
deep generative models. Our method is based on the recently proposed Deep Image
Prior (DIP), wherein the convolutional weights of the network are optimized to
match the observed measurements. We show that this approach can be applied to
solve any differentiable linear inverse problem, outperforming previous
unlearned methods. Unlike various learned approaches based on generative
models, our method does not require pre-training over large datasets. We
further introduce a novel learned regularization technique, which incorporates
prior information on the network weights. This reduces reconstruction error,
especially for noisy measurements. Finally, we prove that, using the DIP
optimization approach, moderately overparameterized single-layer networks can
perfectly fit any signal despite the non-convex nature of the fitting problem.
This theoretical result provides justification for early stopping.
| stat |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.