title
stringlengths 6
244
| abstract
stringlengths 19
5.09k
| label
stringclasses 10
values |
|---|---|---|
The posterior probability of a null hypothesis given a statistically
significant result | When researchers carry out a null hypothesis significance test, it is
tempting to assume that a statistically significant result lowers Prob(H0), the
probability of the null hypothesis being true. Technically, such a statement is
meaningless for various reasons: e.g., the null hypothesis does not have a
probability associated with it. However, it is possible to relax certain
assumptions to compute the posterior probability Prob(H0) under repeated
sampling. We show in a step-by-step guide that the intuitively appealing
belief, that Prob(H0) is low when significant results have been obtained under
repeated sampling, is in general incorrect and depends greatly on: (a) the
prior probability of the null being true; (b) Type I error, and (c) Type II
error. Through step-by-step simulations using open-source code in the R System
of Statistical Computing, we show that uncertainty about the null hypothesis
being true often remains high despite a significant result. To help the reader
develop intuitions about this common misconception, we provide a Shiny app
(https://danielschad.shinyapps.io/probnull/). We expect that this tutorial will
be widely useful for students and researchers in many fields of psychological
science and beyond, and will help researchers better understand and judge
results from null hypothesis significance tests.
| stat |
Fully Bayesian Estimation under Dependent and Informative Cluster
Sampling | Survey data are often collected under multistage sampling designs where units
are binned to clusters that are sampled in a first stage. The unit-indexed
population variables of interest are typically dependent within cluster. We
propose a Fully Bayesian method that constructs an exact likelihood for the
observed sample to incorporate unit-level marginal sampling weights for
performing unbiased inference for population parameters while simultaneously
accounting for the dependence induced by sampling clusters of units to produce
correct uncertainty quantification. Our approach parameterizes cluster-indexed
random effects in both a marginal model for the response and a conditional
model for published, unit-level sampling weights. We compare our method to
plug-in Bayesian and frequentist alternatives in a simulation study and
demonstrate that our method most closely achieves correct uncertainty
quantification for model parameters, including the generating variances for
cluster-indexed random effects. We demonstrate our method in two applications
with NHANES data.
| stat |
cutpointr: Improved Estimation and Validation of Optimal Cutpoints in R | 'Optimal cutpoints' for binary classification tasks are often established by
testing which cutpoint yields the best discrimination, for example the Youden
index, in a specific sample. This results in 'optimal' cutpoints that are
highly variable and systematically overestimate the out-of-sample performance.
To address these concerns, the cutpointr package offers robust methods for
estimating optimal cutpoints and the out-of-sample performance. The robust
methods include bootstrapping and smoothing based on kernel estimation,
generalized additive models, smoothing splines, and local regression. These
methods can be applied to a wide range of binary-classification and cost-based
metrics. cutpointr also provides mechanisms to utilize user-defined metrics and
estimation methods. The package has capabilities for parallelization of the
bootstrapping, including reproducible random number generation. Furthermore, it
is pipe-friendly, for example for compatibility with functions from tidyverse.
Various functions for plotting receiver operating characteristic curves,
precision recall graphs, bootstrap results and other representations of the
data are included. The package contains example data from a study on
psychological characteristics and suicide attempts suitable for applying binary
classification algorithms.
| stat |
Area-level spatio-temporal Poisson mixed models for predicting domain
counts and proportions | This paper introduces area-level Poisson mixed models with temporal and
SAR(1) spatially correlated random effects. Small area predictors of the
proportions and counts of a dichotomic variable are derived from the new models
and the corresponding mean squared errors are estimated by parametric
bootstrap. The paper illustrates the introduced methodology with two
applications to real data. The first one deals with data of forest fires in
Galicia (Spain) during 2007-2008 and the target is modeling and predicting
counts of fires. The second one treats data from the Spanish living conditions
survey of Galicia of 2013 and the target is the estimation of county
proportions of women under the poverty line.
| stat |
Cross-population coupling of neural activity based on Gaussian process
current source densities | Because LFPs arise from multiple sources in different spatial locations, they
do not easily reveal coordinated activity across neural populations on a
trial-to-trial basis. As we show here, however, once disparate source signals
are decoupled, their trial-to-trial fluctuations become more accessible, and
cross-population correlations become more apparent. To decouple sources we
introduce a general framework for estimation of current source densities
(CSDs). In this framework, the set of LFPs result from noise being added to the
transform of the CSD by a biophysical forward model, while the CSD is
considered to be the sum of a zero-mean, stationary, spatiotemporal Gaussian
process, having fast and slow components, and a mean function, which is the sum
of multiple time-varying functions distributed across space, each varying
across trials. We derived biophysical forward models relevant to the data we
analyzed. In simulation studies this approach improved identification of source
signals compared to existing CSD estimation methods. Using data recorded from
primate auditory cortex, we analyzed trial-to-trial fluctuations in both
steady-state and task-evoked signals. We found cortical layer-specific phase
coupling between two probes and showed that the same analysis applied directly
to LFPs did not recover these patterns. We also found task-evoked CSDs to be
correlated across probes, at specific cortical depths. Using data from
Neuropixels probes in mouse visual areas, we again found evidence for
depth-specific phase coupling of areas V1 and LM based on the CSDs.
| stat |
Optimal block designs for experiments on networks | We propose a method for constructing optimal block designs for experiments on
networks. The response model for a given network interference structure extends
the linear network effects model to incorporate blocks. The optimality criteria
are chosen to reflect the experimental objectives and an exchange algorithm is
used to search across the design space for obtaining an efficient design when
an exhaustive search is not possible. Our interest lies in estimating the
direct comparisons among treatments, in the presence of nuisance network
effects that stem from the underlying network interference structure governing
the experimental units, or in the network effects themselves. Comparisons of
optimal designs under different models, including the standard treatment
models, are examined by comparing the variance and bias of treatment effect
estimators. We also suggest a way of defining blocks, while taking into account
the interrelations of groups of experimental units within a network, using
spectral clustering techniques to achieve optimal modularity. We expect
connected units within closed-form communities to behave similarly to an
external stimulus. We provide evidence that our approach can lead to efficiency
gains over conventional designs such as randomized designs that ignore the
network structure and we illustrate its usefulness for experiments on networks.
| stat |
The Importance of Being Correlated: Implications of Dependence in Joint
Spectral Inference across Multiple Networks | Spectral inference on multiple networks is a rapidly-developing subfield of
graph statistics. Recent work has demonstrated that joint, or simultaneous,
spectral embedding of multiple independent network realizations can deliver
more accurate estimation than individual spectral decompositions of those same
networks. Little attention has been paid, however, to the network correlation
that such joint embedding procedures necessarily induce. In this paper, we
present a detailed analysis of induced correlation in a {\em generalized
omnibus} embedding for multiple networks. We show that our embedding procedure
is flexible and robust, and, moreover, we prove a central limit theorem for
this embedding and explicitly compute the limiting covariance. We examine how
this covariance can impact inference in a network time series, and we construct
an appropriately calibrated omnibus embedding that can detect changes in real
biological networks that previous embedding procedures could not discern. Our
analysis confirms that the effect of induced correlation can be both subtle and
transformative, with import in theory and practice.
| stat |
Learning to Advertise for Organic Traffic Maximization in E-Commerce
Product Feeds | Most e-commerce product feeds provide blended results of advertised products
and recommended products to consumers. The underlying advertising and
recommendation platforms share similar if not exactly the same set of candidate
products. Consumers' behaviors on the advertised results constitute part of the
recommendation model's training data and therefore can influence the
recommended results. We refer to this process as Leverage. Considering this
mechanism, we propose a novel perspective that advertisers can strategically
bid through the advertising platform to optimize their recommended organic
traffic. By analyzing the real-world data, we first explain the principles of
Leverage mechanism, i.e., the dynamic models of Leverage. Then we introduce a
novel Leverage optimization problem and formulate it with a Markov Decision
Process. To deal with the sample complexity challenge in model-free
reinforcement learning, we propose a novel Hybrid Training Leverage Bidding
(HTLB) algorithm which combines the real-world samples and the
emulator-generated samples to boost the learning speed and stability. Our
offline experiments as well as the results from the online deployment
demonstrate the superior performance of our approach.
| stat |
Estimating the parameters of ocean wave spectra | Wind-generated waves are often treated as stochastic processes. There is
particular interest in their spectral density functions, which are often
expressed in some parametric form. Such spectral density functions are used as
inputs when modelling structural response or other engineering concerns.
Therefore, accurate and precise recovery of the parameters of such a form, from
observed wave records, is important. Current techniques are known to struggle
with recovering certain parameters, especially the peak enhancement factor and
spectral tail decay. We introduce an approach from the statistical literature,
known as the de-biased Whittle likelihood, and address some practical concerns
regarding its implementation in the context of wind-generated waves. We
demonstrate, through numerical simulation, that the de-biased Whittle
likelihood outperforms current techniques, such as least squares fitting, both
in terms of accuracy and precision of the recovered parameters. We also provide
a method for estimating the uncertainty of parameter estimates. We perform an
example analysis on a data-set recorded off the coast of New Zealand, to
illustrate some of the extra practical concerns that arise when estimating the
parameters of spectra from observed data.
| stat |
A Robust Consistent Information Criterion for Model Selection based on
Empirical Likelihood | Conventional likelihood-based information criteria for model selection rely
on the distribution assumption of data. However, for complex data that are
increasingly available in many scientific fields, the specification of their
underlying distribution turns out to be challenging, and the existing criteria
may be limited and are not general enough to handle a variety of model
selection problems. Here, we propose a robust and consistent model selection
criterion based upon the empirical likelihood function which is data-driven. In
particular, this framework adopts plug-in estimators that can be achieved by
solving external estimating equations, not limited to the empirical likelihood,
which avoids potential computational convergence issues and allows versatile
applications, such as generalized linear models, generalized estimating
equations, penalized regressions and so on. The formulation of our proposed
criterion is initially derived from the asymptotic expansion of the marginal
likelihood under variable selection framework, but more importantly, the
consistent model selection property is established under a general context.
Extensive simulation studies confirm the out-performance of the proposal
compared to traditional model selection criteria. Finally, an application to
the Atherosclerosis Risk in Communities Study illustrates the practical value
of this proposed framework.
| stat |
Simultaneous Non-Gaussian Component Analysis (SING) for Data Integration
in Neuroimaging | As advances in technology allow the acquisition of complementary information,
it is increasingly common for scientific studies to collect multiple datasets.
Large-scale neuroimaging studies often include multiple modalities (e.g., task
functional MRI, resting-state fMRI, diffusion MRI, and/or structural MRI), with
the aim to understand the relationships between datasets. In this study, we
seek to understand whether regions of the brain activated in a working memory
task relate to resting-state correlations. In neuroimaging, a popular approach
uses principal component analysis for dimension reduction prior to canonical
correlation analysis with joint independent component analysis, but this may
discard biological features with low variance and/or spuriously associate
structure unique to a dataset with joint structure. We introduce Simultaneous
Non-Gaussian component analysis (SING) in which dimension reduction and feature
extraction are achieved simultaneously, and shared information is captured via
subject scores. We apply our method to a working memory task and resting-state
correlations from the Human Connectome Project. We find joint structure as
evident from joint scores whose loadings highlight resting-state correlations
involving regions associated with working memory. Moreover, some of the subject
scores are related to fluid intelligence.
| stat |
Spatial Bayesian Hierarchical Modelling with Integrated Nested Laplace
Approximation | We consider latent Gaussian fields for modelling spatial dependence in the
context of both spatial point patterns and areal data, providing two different
applications. The inhomogeneous Log-Gaussian Cox Process model is specified to
describe a seismic sequence occurred in Greece, resorting to the Stochastic
Partial Differential Equations. The Besag-York-Mollie model is fitted for
disease mapping of the Covid-19 infection in the North of Italy. These models
both belong to the class of Bayesian hierarchical models with latent Gaussian
fields whose posterior is not available in closed form. Therefore, the
inference is performed with the Integrated Nested Laplace Approximation, which
provides accurate and relatively fast analytical approximations to the
posterior quantities of interest.
| stat |
Bayesian forecasting of multivariate time series: Scalability, structure
uncertainty and decisions | I overview recent research advances in Bayesian state-space modeling of
multivariate time series. A main focus is on the decouple/recouple concept that
enables application of state-space models to increasingly large-scale data,
applying to continuous or discrete time series outcomes. The scope includes
large-scale dynamic graphical models for forecasting and multivariate
volatility analysis in areas such as economics and finance, multi-scale
approaches for forecasting discrete/count time series in areas such as
commercial sales and demand forecasting, and dynamic network flow models for
areas including internet traffic monitoring. In applications, explicit
forecasting, monitoring and decision goals are paramount and should factor into
model assessment and comparison, a perspective that is highlighted.
| stat |
Scalable Feature Matching Across Large Data Collections | This paper is concerned with matching feature vectors in a one-to-one fashion
across large collections of datasets. Formulating this task as a
multidimensional assignment problem with decomposable costs (MDADC), we develop
extremely fast algorithms with time complexity linear in the number $n$ of
datasets and space complexity a small fraction of the data size. These
remarkable properties hinge on using the squared Euclidean distance as
dissimilarity function, which can reduce ${n \choose 2}$ matching problems
between pairs of datasets to $n$ problems and enable calculating assignment
costs on the fly. To our knowledge, no other method applicable to the MDADC
possesses these linear scaling and low-storage properties necessary to
large-scale applications. In numerical experiments, the novel algorithms
outperform competing methods and show excellent computational and optimization
performances. An application of feature matching to a large neuroimaging
database is presented. The algorithms of this paper are implemented in the R
package matchFeat available at https://github.com/ddegras/matchFeat.
| stat |
Bounds on Bayes Factors for Binomial A/B Testing | Bayes factors, in many cases, have been proven to bridge the classic -value
based significance testing and bayesian analysis of posterior odds. This paper
discusses this phenomena within the binomial A/B testing setup (applicable for
example to conversion testing). It is shown that the bayes factor is controlled
by the \emph{Jensen-Shannon divergence} of success ratios in two tested groups,
which can be further bounded by the Welch statistic. As a result, bayesian
sample bounds almost match frequentionist's sample bounds. The link between
Jensen-Shannon divergence and Welch's test as well as the derivation are an
elegant application of tools from information geometry.
| stat |
A Higher-Order Kolmogorov-Smirnov Test | We present an extension of the Kolmogorov-Smirnov (KS) two-sample test, which
can be more sensitive to differences in the tails. Our test statistic is an
integral probability metric (IPM) defined over a higher-order total variation
ball, recovering the original KS test as its simplest case. We give an exact
representer result for our IPM, which generalizes the fact that the original KS
test statistic can be expressed in equivalent variational and CDF forms. For
small enough orders ($k \leq 5$), we develop a linear-time algorithm for
computing our higher-order KS test statistic; for all others ($k \geq 6$), we
give a nearly linear-time approximation. We derive the asymptotic null
distribution for our test, and show that our nearly linear-time approximation
shares the same asymptotic null. Lastly, we complement our theory with
numerical studies.
| stat |
Towards Robust Classification with Deep Generative Forests | Decision Trees and Random Forests are among the most widely used machine
learning models, and often achieve state-of-the-art performance in tabular,
domain-agnostic datasets. Nonetheless, being primarily discriminative models
they lack principled methods to manipulate the uncertainty of predictions. In
this paper, we exploit Generative Forests (GeFs), a recent class of deep
probabilistic models that addresses these issues by extending Random Forests to
generative models representing the full joint distribution over the feature
space. We demonstrate that GeFs are uncertainty-aware classifiers, capable of
measuring the robustness of each prediction as well as detecting
out-of-distribution samples.
| stat |
Finding archetypal patterns for binary questionnaires | Archetypal analysis is an exploratory tool that explains a set of
observations as mixtures of pure (extreme) patterns. If the patterns are actual
observations of the sample, we refer to them as archetypoids. For the first
time, we propose to use archetypoid analysis for binary observations. This tool
can contribute to the understanding of a binary data set, as in the
multivariate case. We illustrate the advantages of the proposed methodology in
a simulation study and two applications, one exploring objects (rows) and the
other exploring items (columns). One is related to determining student skill
set profiles and the other to describing item response functions.
| stat |
MAP Clustering under the Gaussian Mixture Model via Mixed Integer
Nonlinear Optimization | We present a global optimization approach for solving the maximum
a-posteriori (MAP) clustering problem under the Gaussian mixture model.Our
approach can accommodate side constraints and it preserves the combinatorial
structure of the MAP clustering problem by formulating it asa mixed-integer
nonlinear optimization problem (MINLP). We approximate the MINLP through a
mixed-integer quadratic program (MIQP) transformation that improves
computational aspects while guaranteeing $\epsilon$-global optimality. An
important benefit of our approach is the explicit quantification of the degree
of suboptimality, via the optimality gap, en route to finding the globally
optimal MAP clustering. Numerical experiments comparing our method to other
approaches show that our method finds a better solution than standard
clustering methods. Finally, we cluster a real breast cancer gene expression
data set incorporating intrinsic subtype information; the induced constraints
substantially improve the computational performance and produce more coherent
and bio-logically meaningful clusters.
| stat |
Clustering and Prediction with Variable Dimension Covariates | In many applied fields incomplete covariate vectors are commonly encountered.
It is well known that this can be problematic when making inference on model
parameters, but its impact on prediction performance is less understood. We
develop a method based on covariate dependent partition models that seamlessly
handles missing covariates while completely avoiding any type of imputation.
The method we develop allows in-sample predictions as well as out-of-sample
prediction, even if the missing pattern in the new subjects' incomplete
covariate vector was not seen in the training data. Any data type, including
categorical or continuous covariates are permitted. In simulation studies the
proposed method compares favorably. We illustrate the method in two application
examples.
| stat |
Network-based models for social recommender systems | With the overwhelming online products available in recent years, there is an
increasing need to filter and deliver relevant personalized advice for users.
Recommender systems solve this problem by modeling and predicting individual
preferences for a great variety of items such as movies, books or research
articles. In this chapter, we explore rigorous network-based models that
outperform leading approaches for recommendation. The network models we
consider are based on the explicit assumption that there are groups of
individuals and of items, and that the preferences of an individual for an item
are determined only by their group memberships. The accurate prediction of
individual user preferences over items can be accomplished by different
methodologies, such as Monte Carlo sampling or Expectation-Maximization
methods, the latter resulting in a scalable algorithm which is suitable for
large datasets.
| stat |
Toward Universal Testing of Dynamic Network Models | Numerous networks in the real world change over time, in the sense that nodes
and edges enter and leave the networks. Various dynamic random graph models
have been proposed to explain the macroscopic properties of these systems and
to provide a foundation for statistical inferences and predictions. It is of
interest to have a rigorous way to determine how well these models match
observed networks. We thus ask the following goodness of fit question: given a
sequence of observations/snapshots of a growing random graph, along with a
candidate model M, can we determine whether the snapshots came from M or from
some arbitrary alternative model that is well-separated from M in some natural
metric? We formulate this problem precisely and boil it down to goodness of fit
testing for graph-valued, infinite-state Markov processes and exhibit and
analyze a universal test based on non-stationary sampling for a natural class
of models.
| stat |
Inference After Selecting Plausibly Valid Instruments with Application
to Mendelian Randomization | Mendelian randomization (MR) is a popular method in genetic epidemiology to
estimate the effect of an exposure on an outcome by using genetic instruments.
These instruments are often selected from a combination of prior knowledge from
genome wide association studies (GWAS) and data-driven instrument selection
procedures or tests. Unfortunately, when testing for the exposure effect, the
instrument selection process done a priori is not accounted for. This paper
studies and highlights the bias resulting from not accounting for the
instrument selection process by focusing on a recent data-driven instrument
selection procedure, sisVIVE, as an example. We introduce a conditional
inference approach that conditions on the instrument selection done a priori
and leverage recent advances in selective inference to derive conditional null
distributions of popular test statistics for the exposure effect in MR. The
null distributions can be characterized with individual-level or summary-level
data in MR. We show that our conditional confidence intervals derived from
conditional null distributions attain the desired nominal level while typical
confidence intervals computed in MR do not. We conclude by reanalyzing the
effect of BMI on diastolic blood pressure using summary-level data from the
UKBiobank that accounts for instrument selection.
| stat |
Using Multiple Pre-treatment Periods to Improve
Difference-in-Differences and Staggered Adoption Design | While difference-in-differences (DID) was originally developed with one pre-
and one post-treatment periods, data from additional pre-treatment periods is
often available. How can researchers improve the DID design with such multiple
pre-treatment periods under what conditions? We first use potential outcomes to
clarify three benefits of multiple pre-treatment periods: (1) assessing the
parallel trends assumption, (2) improving estimation accuracy, and (3) allowing
for a more flexible parallel trends assumption. We then propose a new
estimator, double DID, which combines all the benefits through the generalized
method of moments and contains the two-way fixed effects regression as a
special case. In a wide range of applications where several pre-treatment
periods are available, the double DID improves upon the standard DID both in
terms of identification and estimation accuracy. We also generalize the double
DID to the staggered adoption design where different units can receive the
treatment in different time periods. We illustrate the proposed method with two
empirical applications, covering both the basic DID and staggered adoption
designs. We offer an open-source R package that implements the proposed
methodologies.
| stat |
Feature Selection for multi-labeled variables via Dependency
Maximization | Feature selection and reducing the dimensionality of data is an essential
step in data analysis. In this work, we propose a new criterion for feature
selection that is formulated as conditional information between features given
the labeled variable. Instead of using the standard mutual information measure
based on Kullback-Leibler divergence, we use our proposed criterion to filter
out redundant features for the purpose of multiclass classification. This
approach results in an efficient and fast non-parametric implementation of
feature selection as it can be directly estimated using a geometric measure of
dependency, the global Friedman-Rafsky (FR) multivariate run test statistic
constructed by a global minimal spanning tree (MST). We demonstrate the
advantages of our proposed feature selection approach through simulation. In
addition the proposed feature selection method is applied to the MNIST data
set.
| stat |
Clustering and Trend Analysis of Global Extreme Droughts from 1900 to
2014 | Drought is one of the most devastating environmental disasters. Analyzing
historical changes in climate extremes is critical for mitigating its adverse
impacts in the future. In the present study, the spatial and temporal
characteristics of the global severe droughts using Palmer Drought Intensity
Index (PDSI) from 1900 to 2014 are explored. K-means clustering is implemented
to partition the extreme negative PDSI values. The global extreme droughts
magnitude around the world from 1950 to 1980 were less intense compared to the
other decades. In 2012, the largest areas around the world, especially Canada,
experienced their most severe historical droughts. Results show that the most
recent extreme droughts occurred in some regions such as the North of Canada,
central regions of the US, Southwest of Europe and Southeast Asia. We found
that after 1980, the spatial extent of the regions that experienced extreme
drought have increased substantially.
| stat |
Multitask learning and benchmarking with clinical time series data | Health care is one of the most exciting frontiers in data mining and machine
learning. Successful adoption of electronic health records (EHRs) created an
explosion in digital clinical data available for analysis, but progress in
machine learning for healthcare research has been difficult to measure because
of the absence of publicly available benchmark data sets. To address this
problem, we propose four clinical prediction benchmarks using data derived from
the publicly available Medical Information Mart for Intensive Care (MIMIC-III)
database. These tasks cover a range of clinical problems including modeling
risk of mortality, forecasting length of stay, detecting physiologic decline,
and phenotype classification. We propose strong linear and neural baselines for
all four tasks and evaluate the effect of deep supervision, multitask training
and data-specific architectural modifications on the performance of neural
models.
| stat |
Measuring Sample Quality with Stein's Method | To improve the efficiency of Monte Carlo estimation, practitioners are
turning to biased Markov chain Monte Carlo procedures that trade off asymptotic
exactness for computational speed. The reasoning is sound: a reduction in
variance due to more rapid sampling can outweigh the bias introduced. However,
the inexactness creates new challenges for sampler and parameter selection,
since standard measures of sample quality like effective sample size do not
account for asymptotic bias. To address these challenges, we introduce a new
computable quality measure based on Stein's method that quantifies the maximum
discrepancy between sample and target expectations over a large class of test
functions. We use our tool to compare exact, biased, and deterministic sample
sequences and illustrate applications to hyperparameter selection, convergence
rate assessment, and quantifying bias-variance tradeoffs in posterior
inference.
| stat |
Robust Matrix Completion with Mixed Data Types | We consider the matrix completion problem of recovering a structured low rank
matrix with partially observed entries with mixed data types. Vast majority of
the solutions have proposed computationally feasible estimators with strong
statistical guarantees for the case where the underlying distribution of data
in the matrix is continuous. A few recent approaches have extended using
similar ideas these estimators to the case where the underlying distributions
belongs to the exponential family. Most of these approaches assume that there
is only one underlying distribution and the low rank constraint is regularized
by the matrix Schatten Norm. We propose a computationally feasible statistical
approach with strong recovery guarantees along with an algorithmic framework
suited for parallelization to recover a low rank matrix with partially observed
entries for mixed data types in one step. We also provide extensive simulation
evidence that corroborate our theoretical results.
| stat |
Meta-strategy for Learning Tuning Parameters with Guarantees | Online gradient methods, like the online gradient algorithm (OGA), often
depend on tuning parameters that are difficult to set in practice. We consider
an online meta-learning scenario, and we propose a meta-strategy to learn these
parameters from past tasks. Our strategy is based on the minimization of a
regret bound. It allows to learn the initialization and the step size in OGA
with guarantees. We provide a regret analysis of the strategy in the case of
convex losses. It suggests that, when there are parameters
$\theta_1,\dots,\theta_T$ solving well tasks $1,\dots,T$ respectively and that
are close enough one to each other, our strategy indeed improves on learning
each task in isolation.
| stat |
Non-Negative Networks Against Adversarial Attacks | Adversarial attacks against neural networks are a problem of considerable
importance, for which effective defenses are not yet readily available. We make
progress toward this problem by showing that non-negative weight constraints
can be used to improve resistance in specific scenarios. In particular, we show
that they can provide an effective defense for binary classification problems
with asymmetric cost, such as malware or spam detection. We also show the
potential for non-negativity to be helpful to non-binary problems by applying
it to image classification.
| stat |
Sylvester Normalizing Flows for Variational Inference | Variational inference relies on flexible approximate posterior distributions.
Normalizing flows provide a general recipe to construct flexible variational
posteriors. We introduce Sylvester normalizing flows, which can be seen as a
generalization of planar flows. Sylvester normalizing flows remove the
well-known single-unit bottleneck from planar flows, making a single
transformation much more flexible. We compare the performance of Sylvester
normalizing flows against planar flows and inverse autoregressive flows and
demonstrate that they compare favorably on several datasets.
| stat |
Doubly Robust Inference when Combining Probability and Non-probability
Samples with High-dimensional Data | Non-probability samples become increasingly popular in survey statistics but
may suffer from selection biases that limit the generalizability of results to
the target population. We consider integrating a non-probability sample with a
probability sample which provides high-dimensional representative covariate
information of the target population. We propose a two-step approach for
variable selection and finite population inference. In the first step, we use
penalized estimating equations with folded-concave penalties to select
important variables for the sampling score of selection into the
non-probability sample and the outcome model. We show that the penalized
estimating equation approach enjoys the selection consistency property for
general probability samples. The major technical hurdle is due to the possible
dependence of the sample under the finite population framework. To overcome
this challenge, we construct martingales which enable us to apply Bernstein
concentration inequality for martingales. In the second step, we focus on a
doubly robust estimator of the finite population mean and re-estimate the
nuisance model parameters by minimizing the asymptotic squared bias of the
doubly robust estimator. This estimating strategy mitigates the possible
first-step selection error and renders the doubly robust estimator root-n
consistent if either the sampling probability or the outcome model is correctly
specified.
| stat |
Minimum weight norm models do not always generalize well for
over-parameterized problems | This work is substituted by the paper in arXiv:2011.14066.
Stochastic gradient descent is the de facto algorithm for training deep
neural networks (DNNs). Despite its popularity, it still requires fine tuning
in order to achieve its best performance. This has led to the development of
adaptive methods, that claim automatic hyper-parameter optimization. Recently,
researchers have studied both algorithmic classes via toy examples: e.g., for
over-parameterized linear regression, Wilson et. al. (2017) shows that, while
SGD always converges to the minimum-norm solution, adaptive methods show no
such inclination, leading to worse generalization capabilities. Our aim is to
study this conjecture further. We empirically show that the minimum weight norm
is not necessarily the proper gauge of good generalization in simplified
scenaria, and different models found by adaptive methods could outperform plain
gradient methods. In practical DNN settings, we observe that adaptive methods
can outperform SGD, with larger weight norm output models, but without
necessarily reducing the amount of tuning required.
| stat |
Tensor Programs II: Neural Tangent Kernel for Any Architecture | We prove that a randomly initialized neural network of *any architecture* has
its Tangent Kernel (NTK) converge to a deterministic limit, as the network
widths tend to infinity. We demonstrate how to calculate this limit. In prior
literature, the heuristic study of neural network gradients often assumes every
weight matrix used in forward propagation is independent from its transpose
used in backpropagation (Schoenholz et al. 2017). This is known as the
*gradient independence assumption (GIA)*. We identify a commonly satisfied
condition, which we call *Simple GIA Check*, such that the NTK limit
calculation based on GIA is correct. Conversely, when Simple GIA Check fails,
we show GIA can result in wrong answers. Our material here presents the NTK
results of Yang (2019a) in a friendly manner and showcases the *tensor
programs* technique for understanding wide neural networks. We provide
reference implementations of infinite-width NTKs of recurrent neural network,
transformer, and batch normalization at https://github.com/thegregyang/NTK4A.
| stat |
Physically-Inspired Gaussian Process Models for Post-Transcriptional
Regulation in Drosophila | The regulatory process of Drosophila is thoroughly studied for understanding
a great variety of biological principles. While pattern-forming gene networks
are analysed in the transcription step, post-transcriptional events (e.g.
translation, protein processing) play an important role in establishing protein
expression patterns and levels. Since the post-transcriptional regulation of
Drosophila depends on spatiotemporal interactions between mRNAs and gap
proteins, proper physically-inspired stochastic models are required to study
the link between both quantities. Previous research attempts have shown that
using Gaussian processes (GPs) and differential equations lead to promising
predictions when analysing regulatory networks. Here we aim at further
investigating two types of physically-inspired GP models based on a
reaction-diffusion equation where the main difference lies in where the prior
is placed. While one of them has been studied previously using protein data
only, the other is novel and yields a simple approach requiring only the
differentiation of kernel functions. In contrast to other stochastic
frameworks, discretising the spatial space is not required here. Both GP models
are tested under different conditions depending on the availability of gap gene
mRNA expression data. Finally, their performances are assessed on a
high-resolution dataset describing the blastoderm stage of the early embryo of
Drosophila melanogaster
| stat |
Bayesian Image Classification with Deep Convolutional Gaussian Processes | In decision-making systems, it is important to have classifiers that have
calibrated uncertainties, with an optimisation objective that can be used for
automated model selection and training. Gaussian processes (GPs) provide
uncertainty estimates and a marginal likelihood objective, but their weak
inductive biases lead to inferior accuracy. This has limited their
applicability in certain tasks (e.g. image classification). We propose a
translation-insensitive convolutional kernel, which relaxes the translation
invariance constraint imposed by previous convolutional GPs. We show how we can
use the marginal likelihood to learn the degree of insensitivity. We also
reformulate GP image-to-image convolutional mappings as multi-output GPs,
leading to deep convolutional GPs. We show experimentally that our new kernel
improves performance in both single-layer and deep models. We also demonstrate
that our fully Bayesian approach improves on dropout-based Bayesian deep
learning methods in terms of uncertainty and marginal likelihood estimates.
| stat |
Generative Modeling with Denoising Auto-Encoders and Langevin Sampling | We study convergence of a generative modeling method that first estimates the
score function of the distribution using Denoising Auto-Encoders (DAE) or
Denoising Score Matching (DSM) and then employs Langevin diffusion for
sampling. We show that both DAE and DSM provide estimates of the score of the
Gaussian smoothed population density, allowing us to apply the machinery of
Empirical Processes.
We overcome the challenge of relying only on $L^2$ bounds on the score
estimation error and provide finite-sample bounds in the Wasserstein distance
between the law of the population distribution and the law of this sampling
scheme. We then apply our results to the homotopy method of arXiv:1907.05600
and provide theoretical justification for its empirical success.
| stat |
Active embedding search via noisy paired comparisons | Suppose that we wish to estimate a user's preference vector $w$ from paired
comparisons of the form "does user $w$ prefer item $p$ or item $q$?," where
both the user and items are embedded in a low-dimensional Euclidean space with
distances that reflect user and item similarities. Such observations arise in
numerous settings, including psychometrics and psychology experiments, search
tasks, advertising, and recommender systems. In such tasks, queries can be
extremely costly and subject to varying levels of response noise; thus, we aim
to actively choose pairs that are most informative given the results of
previous comparisons. We provide new theoretical insights into the benefits and
challenges of greedy information maximization in this setting, and develop two
novel strategies that maximize lower bounds on information gain and are simpler
to analyze and compute respectively. We use simulated responses from a
real-world dataset to validate our strategies through their similar performance
to greedy information maximization, and their superior preference estimation
over state-of-the-art selection methods as well as random queries.
| stat |
Bayesian Beta-Binomial Prevalence Estimation Using an Imperfect Test | Following [Diggle 2011, Greenland 1995], we give a simple formula for the
Bayesian posterior density of a prevalence parameter based on unreliable
testing of a population. This problem is of particular importance when the
false positive test rate is close to the prevalence in the population being
tested. An efficient Monte Carlo algorithm for approximating the posterior
density is presented, and applied to estimating the Covid-19 infection rate in
Santa Clara county, CA using the data reported in [Bendavid 2020]. We show that
the true Bayesian posterior places considerably more mass near zero, resulting
in a prevalence estimate of 5,000--70,000 infections (median: 42,000) (2.17%
(95CI 0.27%--3.63%)), compared to the estimate of 48,000--81,000 infections
derived in [Bendavid 2020] using the delta method.
A demonstration, with code and additional examples, is available at
testprev.com.
| stat |
Generalized Maximum Entropy for Supervised Classification | The maximum entropy principle advocates to evaluate events' probabilities
using a distribution that maximizes entropy among those that satisfy certain
expectations' constraints. Such principle can be generalized for arbitrary
decision problems where it corresponds to minimax approaches. This paper
establishes a framework for supervised classification based on the generalized
maximum entropy principle that leads to minimax risk classifiers (MRCs). We
develop learning techniques that determine MRCs for general entropy functions
and provide performance guarantees by means of convex optimization. In
addition, we describe the relationship of the presented techniques with
existing classification methods, and quantify MRCs performance in comparison
with the proposed bounds and conventional methods.
| stat |
Break Point Detection for Functional Covariance | Many experiments record sequential trajectories that oscillate around zero.
Such trajectories can be viewed as zero-mean functional data. When there are
structural breaks (on the sequence of curves) in higher order moments, it is
often difficult to spot these by mere visual inspection. Thus, we propose a
detection and testing procedure to find the change-points in functional
covariance. The method is fully functional in the sense that no dimension
reduction is needed. We establish the asymptotic properties of the estimated
change-point. The effectiveness of the proposed method is numerically validated
in the simulation studies and an application to study structural changes in rat
brain signals in a stroke experiment.
| stat |
A practical test for a planted community in heterogeneous networks | One of the fundamental task in graph data mining is to find a planted
community(dense subgraph), which has wide application in biology, finance, spam
detection and so on. For a real network data, the existence of a dense subgraph
is generally unknown. Statistical tests have been devised to testing the
existence of dense subgraph in a homogeneous random graph. However, many
networks present extreme heterogeneity, that is, the degrees of nodes or
vertexes don't concentrate on a typical value. The existing tests designed for
homogeneous random graph are not straightforwardly applicable to the
heterogeneous case. Recently, scan test was proposed for detecting a dense
subgraph in heterogeneous(inhomogeneous) graph(\cite{BCHV19}). However, the
computational complexity of the scan test is generally not polynomial in the
graph size, which makes the test impractical for large or moderate networks. In
this paper, we propose a polynomial-time test that has the standard normal
distribution as the null limiting distribution. The power of the test is
theoretically investigated and we evaluate the performance of the test by
simulation and real data example.
| stat |
Modeling and Forecasting Art Movements with CGANs | Conditional Generative Adversarial Networks~(CGAN) are a recent and popular
method for generating samples from a probability distribution conditioned on
latent information. The latent information often comes in the form of a
discrete label from a small set. We propose a novel method for training CGANs
which allows us to condition on a sequence of continuous latent distributions
$f^{(1)}, \ldots, f^{(K)}$. This training allows CGANs to generate samples from
a sequence of distributions. We apply our method to paintings from a sequence
of artistic movements, where each movement is considered to be its own
distribution. Exploiting the temporal aspect of the data, a vector
autoregressive (VAR) model is fitted to the means of the latent distributions
that we learn, and used for one-step-ahead forecasting, to predict the latent
distribution of a future art movement $f^{{(K+1)}}$. Realisations from this
distribution can be used by the CGAN to generate "future" paintings. In
experiments, this novel methodology generates accurate predictions of the
evolution of art. The training set consists of a large dataset of past
paintings. While there is no agreement on exactly what current art period we
find ourselves in, we test on plausible candidate sets of present art, and show
that the mean distance to our predictions is small.
| stat |
Supervised Community Detection with Line Graph Neural Networks | Traditionally, community detection in graphs can be solved using spectral
methods or posterior inference under probabilistic graphical models. Focusing
on random graph families such as the stochastic block model, recent research
has unified both approaches and identified both statistical and computational
detection thresholds in terms of the signal-to-noise ratio. By recasting
community detection as a node-wise classification problem on graphs, we can
also study it from a learning perspective. We present a novel family of Graph
Neural Networks (GNNs) for solving community detection problems in a supervised
learning setting. We show that, in a data-driven manner and without access to
the underlying generative models, they can match or even surpass the
performance of the belief propagation algorithm on binary and multi-class
stochastic block models, which is believed to reach the computational
threshold. In particular, we propose to augment GNNs with the non-backtracking
operator defined on the line graph of edge adjacencies. Our models also achieve
good performance on real-world datasets. In addition, we perform the first
analysis of the optimization landscape of training linear GNNs for community
detection problems, demonstrating that under certain simplifications and
assumptions, the loss values at local and global minima are not far apart.
| stat |
Model Selection of Nested and Non-Nested Item Response Models using
Vuong Tests | In this paper, we apply Vuong's (1989) general approach of model selection to
the comparison of nested and non-nested unidimensional and multidimensional
item response theory (IRT) models. Vuong's approach of model selection is
useful because it allows for formal statistical tests of both nested and
non-nested models. However, only the test of non-nested models has been applied
in the context of IRT models to date. After summarizing the statistical theory
underlying the tests, we investigate the performance of all three distinct
Vuong tests in the context of IRT models using simulation studies and real
data. In the non-nested case we observed that the tests can reliably
distinguish between the graded response model and the generalized partial
credit model. In the nested case, we observed that the tests typically perform
as well as or sometimes better than the traditional likelihood ratio test.
Based on these results, we argue that Vuong's approach provides a useful set of
tools for researchers and practitioners to effectively compare competing nested
and non-nested IRT models.
| stat |
Learning RUMs: Reducing Mixture to Single Component via PCA | We consider the problem of learning a mixture of Random Utility Models
(RUMs). Despite the success of RUMs in various domains and the versatility of
mixture RUMs to capture the heterogeneity in preferences, there has been only
limited progress in learning a mixture of RUMs from partial data such as
pairwise comparisons. In contrast, there have been significant advances in
terms of learning a single component RUM using pairwise comparisons. In this
paper, we aim to bridge this gap between mixture learning and single component
learning of RUM by developing a `reduction' procedure. We propose to utilize
PCA-based spectral clustering that simultaneously `de-noises' pairwise
comparison data. We prove that our algorithm manages to cluster the partial
data correctly (i.e., comparisons from the same RUM component are grouped in
the same cluster) with high probability even when data is generated from a
possibly {\em heterogeneous} mixture of well-separated {\em generic} RUMs. Both
the time and the sample complexities scale polynomially in model parameters
including the number of items. Two key features in the analysis are in
establishing (1) a meaningful upper bound on the sub-Gaussian norm for RUM
components embedded into the vector space of pairwise marginals and (2) the
robustness of PCA with missing values in the $L_{2, \infty}$ sense, which might
be of interest in their own right.
| stat |
Bayesian Imaging With Data-Driven Priors Encoded by Neural Networks:
Theory, Methods, and Algorithms | This paper proposes a new methodology for performing Bayesian inference in
imaging inverse problems where the prior knowledge is available in the form of
training data. Following the manifold hypothesis and adopting a generative
modelling approach, we construct a data-driven prior that is supported on a
sub-manifold of the ambient space, which we can learn from the training data by
using a variational autoencoder or a generative adversarial network. We
establish the existence and well-posedness of the associated posterior
distribution and posterior moments under easily verifiable conditions,
providing a rigorous underpinning for Bayesian estimators and uncertainty
quantification analyses. Bayesian computation is performed by using a parallel
tempered version of the preconditioned Crank-Nicolson algorithm on the
manifold, which is shown to be ergodic and robust to the non-convex nature of
these data-driven models. In addition to point estimators and uncertainty
quantification analyses, we derive a model misspecification test to
automatically detect situations where the data-driven prior is unreliable, and
explain how to identify the dimension of the latent space directly from the
training data. The proposed approach is illustrated with a range of experiments
with the MNIST dataset, where it outperforms alternative image reconstruction
approaches from the state of the art. A model accuracy analysis suggests that
the Bayesian probabilities reported by the data-driven models are also
remarkably accurate under a frequentist definition of probability.
| stat |
Robust penalized spline estimation with difference penalties | Penalized spline estimation with discrete difference penalties (P-splines) is
a popular estimation method in semiparametric models, but the classical
least-squares estimator is susceptible to gross errors and other model
deviations. To remedy this deficiency we introduce and study a broad class of
P-spline estimators based on general loss functions. Robust estimators are
obtained by well-chosen loss functions, such as the Huber or Tukey loss
function. A preliminary scale estimator can also be included in the loss
function. We show in this paper that this class of P-spline estimators enjoys
the same optimal asymptotic properties as least-squares P-splines, thereby
providing strong theoretical motivation for its use. The proposed estimators
may be computed very efficiently through a simple adaptation of
well-established iterative least squares algorithms and exhibit excellent
performance even in finite samples, as evidenced by a numerical study and a
real-data example.
| stat |
Continuous-discrete smoothing of diffusions | Suppose X is a multivariate diffusion process that is observed discretely in
time. At each observation time, a linear transformation of the state of the
process is observed with noise. The smoothing problem consists of recovering
the path of the process, consistent with the observations. We derive a novel
Markov Chain Monte Carlo algorithm to sample from the exact smoothing
distribution. The resulting algorithm is called the Backward Filtering Forward
Guiding (BFFG) algorithm. We extend the algorithm to include parameter
estimation. The proposed method relies on guided proposals introduced in
Schauer et al. (2017). We illustrate its efficiency in a number of challenging
problems.
| stat |
A review of Approximate Bayesian Computation methods via density
estimation: inference for simulator-models | This paper provides a review of Approximate Bayesian Computation (ABC)
methods for carrying out Bayesian posterior inference, through the lens of
density estimation. We describe several recent algorithms and make connection
with traditional approaches. We show advantages and limitations of models based
on parametric approaches and we then draw attention to developments in machine
learning, which we believe have the potential to make ABC scalable to higher
dimensions and may be the future direction for research in this area.
| stat |
A Bayesian perspective on sampling of alternatives | In this paper, we apply a Bayesian perspective to sampling of alternatives
for multinomial logit (MNL) and mixed multinomial logit (MMNL) models. We find
three theoretical results -- i) McFadden's correction factor under the uniform
sampling protocol can be transferred to the Bayesian context in MNL; ii) the
uniform sampling protocol minimises the loss in information on the parameters
of interest (i.e. the kernel of the posterior density) and thereby has
desirable small sample properties in MNL; and iii) our theoretical results
extend to Bayesian MMNL models using data augmentation. Notably, sampling of
alternatives in Bayesian MMNL models does not require the inclusion of the
additional correction factor, as identified by Guevara and Ben-Akiva (2013a) in
classical settings. Accordingly, due to desirable small and large sample
properties, uniform sampling is the recommended sampling protocol in MNL and
MMNL, irrespective of the estimation framework selected.
| stat |
Simultaneous Feature Selection and Outlier Detection with Optimality
Guarantees | Sparse estimation methods capable of tolerating outliers have been broadly
investigated in the last decade. We contribute to this research considering
high-dimensional regression problems contaminated by multiple mean-shift
outliers which affect both the response and the design matrix. We develop a
general framework for this class of problems and propose the use of
mixed-integer programming to simultaneously perform feature selection and
outlier detection with provably optimal guarantees. We characterize the
theoretical properties of our approach, i.e. a necessary and sufficient
condition for the robustly strong oracle property, which allows the number of
features to exponentially increase with the sample size; the optimal estimation
of the parameters; and the breakdown point of the resulting estimates.
Moreover, we provide computationally efficient procedures to tune integer
constraints and to warm-start the algorithm. We show the superior performance
of our proposal compared to existing heuristic methods through numerical
simulations and an application investigating the relationships between the
human microbiome and childhood obesity.
| stat |
Logistic Regression with Missing Covariates -- Parameter Estimation,
Model Selection and Prediction within a Joint-Modeling Framework | Logistic regression is a common classification method in supervised learning.
Surprisingly, there are very few solutions for performing logistic regression
with missing values in the covariates. We suggest a complete approach based on
a stochastic approximation version of the EM algorithm to do statistical
inference with missing values including the estimation of the parameters and
their variance, derivation of confidence intervals and a model selection
procedure. We also tackle the problem of prediction for new observations (on a
test set) with missing covariate data. The methodology is computationally
efficient, and its good coverage and variable selection properties are
demonstrated in a simulation study where we contrast its performances to other
methods. For instance, the popular approach of multiple imputation by chained
equations can lead to estimates that exhibit meaningfully greater biases than
the proposed approach. We then illustrate the method on a dataset of severely
traumatized patients from Paris hospitals to predict the occurrence of
hemorrhagic shock, a leading cause of early preventable death in severe trauma
cases. The aim is to consolidate the current red flag procedure, a binary alert
identifying patients with a high risk of severe hemorrhage. The methodology is
implemented in the R package misaem.
| stat |
A Bayesian time-to-event pharmacokinetic model for sequential phase I
dose-escalation trials with multiple schedules | Phase I dose-escalation trials constitute the first step in investigating the
safety of potentially promising drugs in humans. Conventional methods for phase
I dose-escalation trials are based on a single treatment schedule only. More
recently, however, multiple schedules are more frequently investigated in the
same trial. Here, we consider sequential phase I trials, where the trial
proceeds with a new schedule (e.g. daily or weekly dosing) once the dose
escalation with another schedule has been completed. The aim is to utilize the
information from both the completed and the ongoing dose-escalation trial to
inform decisions on the dose level for the next dose cohort. For this purpose,
we adapted the time-to-event pharmacokinetics (TITE-PK) model, which were
originally developed for simultaneous investigation of multiple schedules.
TITE-PK integrates information from multiple schedules using a pharmacokinetics
(PK) model. In a simulation study, the developed appraoch is compared to the
bridging continual reassessment method and the Bayesian logistic regression
model using a meta-analytic-prior. TITE-PK results in better performance than
comparators in terms of recommending acceptable dose and avoiding overly toxic
doses for sequential phase I trials in most of the scenarios considered.
Furthermore, better performance of TITE-PK is achieved while requiring similar
number of patients in the simulated trials. For the scenarios involving one
schedule, TITE-PK displays similar performance with alternatives in terms of
acceptable dose recommendations. The \texttt{R} and \texttt{Stan} code for the
implementation of an illustrative sequential phase I trial example is publicly
available at https://github.com/gunhanb/TITEPK_sequential.
| stat |
Pair-switching rerandomization | Rerandomization discards assignments with covariates unbalanced in the
treatment and control groups to improve the estimation and inference
efficiency. However, the acceptance-rejection sampling method used by
rerandomization is computationally inefficient. As a result, it is
time-consuming for classical rerandomization to draw numerous independent
assignments, which are necessary for constructing Fisher randomization tests.
To address this problem, we propose a pair-switching rerandomization method to
draw balanced assignments much efficiently. We show that the
difference-in-means estimator is unbiased for the average treatment effect and
the Fisher randomization tests are valid under pair-switching rerandomization.
In addition, our method is applicable in both non-sequentially and sequentially
randomized experiments. We conduct comprehensive simulation studies to compare
the finite-sample performances of the proposed method and classical
rerandomization. Simulation results indicate that pair-switching
rerandomization leads to comparable power of Fisher randomization tests and is
4-18 times faster than classical rerandomization. Finally, we apply the
pair-switching rerandomization method to analyze two clinical trial data sets,
both demonstrating the advantages of our method.
| stat |
Adaptive Approximation and Generalization of Deep Neural Network with
Intrinsic Dimensionality | In this study, we prove that an intrinsic low dimensionality of covariates is
the main factor that determines the performance of deep neural networks (DNNs).
DNNs generally provide outstanding empirical performance. Hence, numerous
studies have actively investigated the theoretical properties of DNNs to
understand their underlying mechanisms. In particular, the behavior of DNNs in
terms of high-dimensional data is one of the most critical questions. However,
this issue has not been sufficiently investigated from the aspect of
covariates, although high-dimensional data have practically low intrinsic
dimensionality. In this study, we derive bounds for an approximation error and
a generalization error regarding DNNs with intrinsically low dimensional
covariates. We apply the notion of the Minkowski dimension and develop a novel
proof technique. Consequently, we show that convergence rates of the errors by
DNNs do not depend on the nominal high dimensionality of data, but on its lower
intrinsic dimension. We further prove that the rate is optimal in the minimax
sense. We identify an advantage of DNNs by showing that DNNs can handle a
broader class of intrinsic low dimensional data than other adaptive estimators.
Finally, we conduct a numerical simulation to validate the theoretical results.
| stat |
The Political Significance of Social Penumbras | To explain the political clout of different social groups, traditional
accounts typically focus on the group's size, resources, or commonality and
intensity of its members' interests. We contend that a group's "penumbra"-the
set of individuals who are personally familiar with people in that group--is
another important explanatory factor that merits systematic analysis. To this
end, we designed a panel study that allows us to learn about the
characteristics of the penumbras of politically relevant groups such as gay
people, the unemployed or recent immigrants. Our study reveals major and
systematic differences in the penumbras of various social groups, even ones of
similar size. Moreover, we find evidence that entering a group's penumbra is
associated with a change in attitude on related political questions. Taken
together, our findings suggest that penumbras help account for variation in the
political standing of different groups in society.
| stat |
Bayesian spectral density estimation using P-splines with quantile-based
knot placement | This article proposes a Bayesian approach to estimating the spectral density
of a stationary time series using a prior based on a mixture of P-spline
distributions. Our proposal is motivated by the B-spline Dirichlet process
prior of Edwards et al. (2019) in combination with Whittle's likelihood and
aims at reducing the high computational complexity of its posterior
computations. The strength of the B-spline Dirichlet process prior over the
Bernstein-Dirichlet process prior of Choudhuri et al. (2004) lies in its
ability to estimate spectral densities with sharp peaks and abrupt changes due
to the flexibility of B-splines with variable number and location of knots.
Here, we suggest to use P-splines of Eilers and Marx (1996) that combine a
B-spline basis with a discrete penalty on the basis coefficients. In addition
to equidistant knots, a novel strategy for a more expedient placement of knots
is proposed that makes use of the information provided by the periodogram about
the steepness of the spectral power distribution. We demonstrate in a
simulation study and two real case studies that this approach retains the
flexibility of the B-splines, achieves similar ability to accurately estimate
peaks due to the new data-driven knot allocation scheme but significantly
reduces the computational costs.
| stat |
False Discovery Rate Control via Debiased Lasso | We consider the problem of variable selection in high-dimensional statistical
models where the goal is to report a set of variables, out of many predictors
$X_1, \dotsc, X_p$, that are relevant to a response of interest. For linear
high-dimensional model, where the number of parameters exceeds the number of
samples $(p>n)$, we propose a procedure for variables selection and prove that
it controls the "directional" false discovery rate (FDR) below a pre-assigned
significance level $q\in [0,1]$. We further analyze the statistical power of
our framework and show that for designs with subgaussian rows and a common
precision matrix $\Omega\in\mathbb{R}^{p\times p}$, if the minimum nonzero
parameter $\theta_{\min}$ satisfies $$\sqrt{n} \theta_{\min} - \sigma
\sqrt{2(\max_{i\in [p]}\Omega_{ii})\log\left(\frac{2p}{qs_0}\right)} \to
\infty\,,$$ then this procedure achieves asymptotic power one. Our framework is
built upon the debiasing approach and assumes the standard condition $s_0 =
o(\sqrt{n}/(\log p)^2)$, where $s_0$ indicates the number of true positives
among the $p$ features. Notably, this framework achieves exact directional FDR
control without any assumption on the amplitude of unknown regression
parameters, and does not require any knowledge of the distribution of
covariates or the noise level. We test our method in synthetic and real data
experiments to assess its performance and to corroborate our theoretical
results.
| stat |
Transporting a prediction model for use in a new target population | We consider methods for transporting a prediction model and assessing its
performance for use in a new target population, when outcome and covariate
information for model development is available from a simple random sample from
the source population, but only covariate information is available on a simple
random sample from the target population. We discuss how to tailor the
prediction model for use in the target population, how to assess model
performance in the target population (e.g., by estimating the target population
mean squared error), and how to perform model and tuning parameter selection in
the context of the target population. We provide identifiability results for
the target population mean squared error of a potentially misspecified
prediction model under a sampling design where the source study and the target
population samples are obtained separately. We also introduce the concept of
prediction error modifiers that can be used to reason about the need for
tailoring measures of model performance to the target population and provide an
illustration of the methods using simulated data.
| stat |
Variational State-Space Models for Localisation and Dense 3D Mapping in
6 DoF | We solve the problem of 6-DoF localisation and 3D dense reconstruction in
spatial environments as approximate Bayesian inference in a deep state-space
model. Our approach leverages both learning and domain knowledge from
multiple-view geometry and rigid-body dynamics. This results in an expressive
predictive model of the world, often missing in current state-of-the-art visual
SLAM solutions. The combination of variational inference, neural networks and a
differentiable raycaster ensures that our model is amenable to end-to-end
gradient-based optimisation. We evaluate our approach on realistic unmanned
aerial vehicle flight data, nearing the performance of state-of-the-art
visual-inertial odometry systems. We demonstrate the applicability of the model
to generative prediction and planning.
| stat |
Deepening Lee-Carter for longevity projections with uncertainty
estimation | Undoubtedly, several countries worldwide endure to experience a continuous
increase in life expectancy, extending the challenges of life actuaries and
demographers in forecasting mortality. Although several stochastic mortality
models have been proposed in past literature, the mortality forecasting
research remains a crucial task. Recently, various research works encourage the
adequacy of deep learning models to extrapolate suitable pattern within
mortality data. Such a learning models allow to achieve accurate point
predictions, albeit also uncertainty measures are necessary to support both
model estimates reliability and risk evaluations. To the best of our knowledge,
machine and deep learning literature in mortality forecasting lack for studies
about uncertainty estimation. As new advance in mortality forecasting, we
formalizes the deep Neural Networks integration within the Lee-Carter
framework, posing a first bridge between the deep learning and the mortality
density forecasts. We test our model proposal in a numerical application
considering three representative countries worldwide and both genders,
scrutinizing two different fitting periods. Exploiting the meaning of both
biological reasonableness and plausibility of forecasts, as well as performance
metrics, our findings confirm the suitability of deep learning models to
improve the predictive capacity of the Lee-Carter model, providing more
reliable mortality boundaries also on the long-run.
| stat |
Efficient Estimation and Evaluation of Prediction Rules in
Semi-Supervised Settings under Stratified Sampling | In many contemporary applications, large amounts of unlabeled data are
readily available while labeled examples are limited. There has been
substantial interest in semi-supervised learning (SSL) which aims to leverage
unlabeled data to improve estimation or prediction. However, current SSL
literature focuses primarily on settings where labeled data is selected
randomly from the population of interest. Non-random sampling, while posing
additional analytical challenges, is highly applicable to many real world
problems. Moreover, no SSL methods currently exist for estimating the
prediction performance of a fitted model under non-random sampling. In this
paper, we propose a two-step SSL procedure for evaluating a prediction rule
derived from a working binary regression model based on the Brier score and
overall misclassification rate under stratified sampling. In step I, we impute
the missing labels via weighted regression with nonlinear basis functions to
account for nonrandom sampling and to improve efficiency. In step II, we
augment the initial imputations to ensure the consistency of the resulting
estimators regardless of the specification of the prediction model or the
imputation model. The final estimator is then obtained with the augmented
imputations. We provide asymptotic theory and numerical studies illustrating
that our proposals outperform their supervised counterparts in terms of
efficiency gain. Our methods are motivated by electronic health records (EHR)
research and validated with a real data analysis of an EHR-based study of
diabetic neuropathy.
| stat |
Local Bures-Wasserstein Transport: A Practical and Fast Mapping
Approximation | Optimal transport (OT)-based methods have a wide range of applications and
have attracted a tremendous amount of attention in recent years. However, most
of the computational approaches of OT do not learn the underlying transport
map. Although some algorithms have been proposed to learn this map, they rely
on kernel-based methods, which makes them prohibitively slow when the number of
samples increases. Here, we propose a way to learn an approximate transport map
and a parametric approximation of the Wasserstein barycenter. We build an
approximated transport mapping by leveraging the closed-form of Gaussian
(Bures-Wasserstein) transport; we compute local transport plans between matched
pairs of the Gaussian components of each density. The learned map generalizes
to out-of-sample examples. We provide experimental results on simulated and
real data, comparing our proposed method with other mapping estimation
algorithms. Preliminary experiments suggest that our proposed method is not
only faster, with a factor 80 overall running time, but it also requires fewer
components than state-of-the-art methods to recover the support of the
barycenter. From a practical standpoint, it is straightforward to implement and
can be used with a conventional machine learning pipeline.
| stat |
Exact Bootstrap and Permutation Distribution of Wins and Losses in a
Hierarchical Trial | Finkelstein-Schoenfeld, Buyse, Pocock, and other authors have developed
generalizations of the Mann-Whitney test that allow for pairwise patient
comparisons to include a hierarchy of measurements. Various authors present
either asymptotic or randomized methods for analyzing the wins. We use graph
theory concepts to derive exact means and variances for the number of wins, as
a replacement for approximate values obtained from bootstrap analysis or random
sampling from the permutation distribution. The time complexity of our
algorithm is $O(N^2)$, where $N$ is the total number of patients. In any
situation where the mean and variance of a bootstrap sample are used to draw
conclusions, our methodology will be faster and more accurate than the
randomized bootstrap or permutation test.
| stat |
User-Friendly Covariance Estimation for Heavy-Tailed Distributions | We offer a survey of recent results on covariance estimation for heavy-tailed
distributions. By unifying ideas scattered in the literature, we propose
user-friendly methods that facilitate practical implementation. Specifically,
we introduce element-wise and spectrum-wise truncation operators, as well as
their $M$-estimator counterparts, to robustify the sample covariance matrix.
Different from the classical notion of robustness that is characterized by the
breakdown property, we focus on the tail robustness which is evidenced by the
connection between nonasymptotic deviation and confidence level. The key
observation is that the estimators needs to adapt to the sample size,
dimensionality of the data and the noise level to achieve optimal tradeoff
between bias and robustness. Furthermore, to facilitate their practical use, we
propose data-driven procedures that automatically calibrate the tuning
parameters. We demonstrate their applications to a series of structured models
in high dimensions, including the bandable and low-rank covariance matrices and
sparse precision matrices. Numerical studies lend strong support to the
proposed methods.
| stat |
Gaussian Process Models with Low-Rank Correlation Matrices for Both
Continuous and Categorical Inputs | We introduce a method that uses low-rank approximations of cross-correlation
matrices in mixed continuous and categorical Gaussian Process models. This new
method -- called Low-Rank Correlation (LRC) -- offers the ability to flexibly
adapt the number of parameters to the problem at hand by choosing an
appropriate rank of the approximation. Furthermore, we present a systematic
approach of defining test functions that can be used for assessing the accuracy
of models or optimization methods that are concerned with both continuous and
categorical inputs. We compare LRC to existing approaches of modeling the
cross-correlation matrix. It turns out that the new approach performs well in
terms of estimation of cross-correlations and response surface prediction.
Therefore, LRC is a flexible and useful addition to existing methods,
especially for increasing numbers of combinations of levels of the categorical
inputs.
| stat |
Semisupervised Clustering by Queries and Locally Encodable Source Coding | Source coding is the canonical problem of data compression in information
theory. In a locally encodable source coding, each compressed bit depends on
only few bits of the input. In this paper, we show that a recently popular
model of semi-supervised clustering is equivalent to locally encodable source
coding. In this model, the task is to perform multiclass labeling of unlabeled
elements. At the beginning, we can ask in parallel a set of simple queries to
an oracle who provides (possibly erroneous) binary answers to the queries. The
queries cannot involve more than two (or a fixed constant number of) elements.
Now the labeling of all the elements (or clustering) must be performed based on
the noisy query answers. The goal is to recover all the correct labelings while
minimizing the number of such queries. The equivalence to locally encodable
source codes leads us to find lower bounds on the number of queries required in
a variety of scenarios. We provide querying schemes based on pairwise `same
cluster' queries - and pairwise AND queries and show provable performance
guarantees for each of the schemes.
| stat |
Knockoffs for the mass: new feature importance statistics with false
discovery guarantees | An important problem in machine learning and statistics is to identify
features that causally affect the outcome. This is often impossible to do from
purely observational data, and a natural relaxation is to identify features
that are correlated with the outcome even conditioned on all other observed
features. For example, we want to identify that smoking really is correlated
with cancer conditioned on demographics. The knockoff procedure is a recent
breakthrough in statistics that, in theory, can identify truly correlated
features while guaranteeing that the false discovery is limited. The idea is to
create synthetic data -- knockoffs -- that captures correlations amongst the
features. However there are substantial computational and practical challenges
to generating and using knockoffs. This paper makes several key advances that
enable knockoff application to be more efficient and powerful. We develop an
efficient algorithm to generate valid knockoffs from Bayesian Networks. Then we
systematically evaluate knockoff test statistics and develop new statistics
with improved power. The paper combines new mathematical guarantees with
systematic experiments on real and synthetic data.
| stat |
Uncovering the Limits of Adversarial Training against Norm-Bounded
Adversarial Examples | Adversarial training and its variants have become de facto standards for
learning robust deep neural networks. In this paper, we explore the landscape
around adversarial training in a bid to uncover its limits. We systematically
study the effect of different training losses, model sizes, activation
functions, the addition of unlabeled data (through pseudo-labeling) and other
factors on adversarial robustness. We discover that it is possible to train
robust models that go well beyond state-of-the-art results by combining larger
models, Swish/SiLU activations and model weight averaging. We demonstrate large
improvements on CIFAR-10 and CIFAR-100 against $\ell_\infty$ and $\ell_2$
norm-bounded perturbations of size $8/255$ and $128/255$, respectively. In the
setting with additional unlabeled data, we obtain an accuracy under attack of
65.88% against $\ell_\infty$ perturbations of size $8/255$ on CIFAR-10 (+6.35%
with respect to prior art). Without additional data, we obtain an accuracy
under attack of 57.20% (+3.46%). To test the generality of our findings and
without any additional modifications, we obtain an accuracy under attack of
80.53% (+7.62%) against $\ell_2$ perturbations of size $128/255$ on CIFAR-10,
and of 36.88% (+8.46%) against $\ell_\infty$ perturbations of size $8/255$ on
CIFAR-100. All models are available at
https://github.com/deepmind/deepmind-research/tree/master/adversarial_robustness.
| stat |
Learning an arbitrary mixture of two multinomial logits | In this paper, we consider mixtures of multinomial logistic models (MNL),
which are known to $\epsilon$-approximate any random utility model. Despite its
long history and broad use, rigorous results are only available for learning a
uniform mixture of two MNLs. Continuing this line of research, we study the
problem of learning an arbitrary mixture of two MNLs. We show that the
identifiability of the mixture models may only fail on an algebraic variety of
a negligible measure. This is done by reducing the problem of learning a
mixture of two MNLs to the problem of solving a system of univariate quartic
equations. We also devise an algorithm to learn any mixture of two MNLs using a
polynomial number of samples and a linear number of queries, provided that a
mixture of two MNLs over some finite universe is identifiable. Several
numerical experiments and conjectures are also presented.
| stat |
Regret Bounds for Noise-Free Bayesian Optimization | Bayesian optimisation is a powerful method for non-convex black-box
optimization in low data regimes. However, the question of establishing tight
upper bounds for common algorithms in the noiseless setting remains a largely
open question. In this paper, we establish new and tightest bounds for two
algorithms, namely GP-UCB and Thompson sampling, under the assumption that the
objective function is smooth in terms of having a bounded norm in a Mat\'ern
RKHS. Importantly, unlike several related works, we do not consider perfect
knowledge of the kernel of the Gaussian process emulator used within the
Bayesian optimization loop. This allows us to provide results for practical
algorithms that sequentially estimate the Gaussian process kernel parameters
from the available data.
| stat |
Approximating exponential family models (not single distributions) with
a two-network architecture | Recently much attention has been paid to deep generative models, since they
have been used to great success for variational inference, generation of
complex data types, and more. In most all of these settings, the goal has been
to find a particular member of that model family: optimized parameters index a
distribution that is close (via a divergence or classification metric) to a
target distribution. Much less attention, however, has been paid to the problem
of learning a model itself. Here we introduce a two-network architecture and
optimization procedure for learning intractable exponential family models (not
a single distribution from those models). These exponential families are
learned accurately, allowing operations like posterior inference to be executed
directly and generically with an input choice of natural parameters, rather
than performing inference via optimization for each particular distribution
within that model.
| stat |
Log-Linear Bayesian Additive Regression Trees for Multinomial Logistic
and Count Regression Models | We introduce Bayesian additive regression trees (BART) for log-linear models
including multinomial logistic regression and count regression with
zero-inflation and overdispersion. BART has been applied to nonparametric mean
regression and binary classification problems in a range of settings. However,
existing applications of BART have been limited to models for Gaussian "data",
either observed or latent. This is primarily because efficient MCMC algorithms
are available for Gaussian likelihoods. But while many useful models are
naturally cast in terms of latent Gaussian variables, many others are not --
including models considered in this paper.
We develop new data augmentation strategies and carefully specified prior
distributions for these new models. Like the original BART prior, the new prior
distributions are carefully constructed and calibrated to be flexible while
guarding against overfitting. Together the new priors and data augmentation
schemes allow us to implement an efficient MCMC sampler outside the context of
Gaussian models. The utility of these new methods is illustrated with examples
and an application to a previously published dataset.
| stat |
Robust Causal Inference Under Covariate Shift via Worst-Case
Subpopulation Treatment Effects | We propose the worst-case treatment effect (WTE) across all subpopulations of
a given size, a conservative notion of topline treatment effect. Compared to
the average treatment effect (ATE), whose validity relies on the covariate
distribution of collected data, WTE is robust to unanticipated covariate
shifts, and positive findings guarantee uniformly valid treatment effects over
subpopulations. We develop a semiparametrically efficient estimator for the
WTE, leveraging machine learning-based estimates of the heterogeneous treatment
effect and propensity score. By virtue of satisfying a key (Neyman)
orthogonality property, our estimator enjoys central limit behavior---oracle
rates with true nuisance parameters---even when estimates of nuisance
parameters converge at slower rates. For both randomized trials and
observational studies, we establish a semiparametric efficiency bound, proving
that our estimator achieves the optimal asymptotic variance. On real datasets
where robustness to covariate shift is of core concern, we illustrate the
non-robustness of ATE under even mild distributional shift, and demonstrate
that the WTE guards against brittle findings that are invalidated by
unanticipated covariate shifts.
| stat |
A hybrid Gibbs sampler for edge-preserving tomographic reconstruction
with uncertain view angles | In computed tomography, data consist of measurements of the attenuation of
X-rays passing through an object. The goal is to reconstruct the linear
attenuation coefficient of the object's interior. For each position of the
X-ray source, characterized by its angle with respect to a fixed coordinate
system, one measures a set of data referred to as a view. A common assumption
is that these view angles are known, but in some applications they are known
with imprecision. We propose a framework to solve a Bayesian inverse problem
that jointly estimates the view angles and an image of the object's attenuation
coefficient. We also include a few hyperparameters that characterize the
likelihood and the priors. Our approach is based on a Gibbs sampler where the
associated conditional densities are simulated using different sampling schemes
- hence the term hybrid. In particular, the conditional distribution associated
with the reconstruction is nonlinear in the image pixels, non-Gaussian and
high-dimensional. We approach this distribution by constructing a Laplace
approximation that represents the target conditional locally at each Gibbs
iteration. This enables sampling of the attenuation coefficients in an
efficient manner using iterative reconstruction algorithms. The numerical
results show that our algorithm is able to jointly identify the image and the
view angles, while also providing uncertainty estimates of both. We demonstrate
our method with 2D X-ray computed tomography problems using fan beam
configurations.
| stat |
Jenss-Bayley Latent Change Score Model with Individual Ratio of Growth
Acceleration in the Framework of Individual Measurement Occasions | Longitudinal analysis has been widely employed to examine between-individual
differences in within-individual change. One challenge for such analyses lies
in that the rate-of-change is only available indirectly when change patterns
are nonlinear with respect to time. Latent change score models (LCSMs), which
can be employed to investigate the change in growth rate at the individual
level, have been developed to address this challenge. We extend an existing
LCSM with the Jenss-Bayley growth curve (Grimm et al., 2016c) and propose a
novel expression of change scores that allows for (1) unequally-spaced study
waves and (2) individual measurement occasions around each wave. We also extend
the existing model to estimate the individual ratio of growth acceleration
(that largely determines the trajectory shape and then is viewed as the most
important parameter in the Jenss-Bayley model). We present the proposed model
by simulation studies and a real-world data analysis. Our simulation studies
demonstrate that the proposed model generally estimates the parameters of
interest unbiasedly, precisely and exhibits appropriate confidence interval
coverage. More importantly, the proposed model with the novel expression of
change scores performed better than the existing model shown by simulation
studies. An empirical example using longitudinal reading scores shows that the
model can estimate the individual ratio of growth acceleration and generate
individual growth rate in practice. We also provide the corresponding code for
the proposed model.
| stat |
Informative extended Mallows priors in the Bayesian Mallows model | The aim of this work is to study the problem of prior elicitation for the
Mallows model with Spearman's distance, a popular distance-based model for
rankings or permutation data. Previous Bayesian inference for such model has
been limited to the use of the uniform prior over the space of permutations. We
present a novel strategy to elicit subjective prior beliefs on the location
parameter of the model, discussing the interpretation of hyper-parameters and
the implication of prior choices for the posterior analysis.
| stat |
Multiple Instance Dictionary Learning for Beat-to-Beat Heart Rate
Monitoring from Ballistocardiograms | A multiple instance dictionary learning approach, Dictionary Learning using
Functions of Multiple Instances (DL-FUMI), is used to perform beat-to-beat
heart rate estimation and to characterize heartbeat signatures from
ballistocardiogram (BCG) signals collected with a hydraulic bed sensor. DL-FUMI
estimates a "heartbeat concept" that represents an individual's personal
ballistocardiogram heartbeat pattern. DL-FUMI formulates heartbeat detection
and heartbeat characterization as a multiple instance learning problem to
address the uncertainty inherent in aligning BCG signals with ground truth
during training. Experimental results show that the estimated heartbeat concept
found by DL-FUMI is an effective heartbeat prototype and achieves superior
performance over comparison algorithms.
| stat |
Parameter Estimation in Abruptly Changing Dynamic Environments | Many real-life dynamical systems change abruptly followed by almost
stationary periods. In this paper, we consider streams of data with such abrupt
behavior and investigate the problem of tracking their statistical properties
in an online manner.
We devise a tracking procedure where an estimator that is suitable for a
stationary environment is combined together with an event detection method such
that the estimator rapidly can jump to a more suitable value if an event is
detected. Combining an estimation procedure with detection procedure is
commonly known idea in the literature. However, our contribution lies in
building the detection procedure based on the difference between the stationary
estimator and a Stochastic Learning Weak Estimator (SLWE). The SLWE estimator
is known to be the state-of-the art approach to tracking properties of
non-stationary environments and thus should be a better choice to detect
changes in abruptly changing environments than the far more common sliding
window based approaches. To the best of our knowledge, the event detection
procedure suggested by Ross et al. (2012) is the only procedure in the
literature taking advantage of the powerful tracking properties of the SLWE
estimator. The procedure in Ross et al. is however quite complex and not well
founded theoretically compared to the procedures in this paper. In this paper,
we focus on estimation procedure for the binomial and multinomial
distributions, but our approach can be easily generalized to cover other
distributions as well.
Extensive simulation results based on both synthetic and real-life data
related to news classification demonstrate that our estimation procedure is
easy to tune and performs well.
| stat |
Testing the equality of multivariate means when $p>n$ by combining the
Hoteling and Simes tests | We propose a method of testing the shift between mean vectors of two
multivariate Gaussian random variables in a high-dimensional setting
incorporating the possible dependency and allowing $p > n$. This method is a
combination of two well-known tests: the Hotelling test and the Simes test. The
tests are integrated by sampling several dimensions at each iteration, testing
each using the Hotelling test, and combining their results using the Simes
test. We prove that this procedure is valid asymptotically. This procedure can
be extended to handle non-equal covariance matrices by plugging in the
appropriate extension of the Hotelling test. Using a simulation study, we show
that the proposed test is advantageous over state-of-the-art tests in many
scenarios and robust to violation of the Gaussian assumption.
| stat |
Deep Structural Causal Models for Tractable Counterfactual Inference | We formulate a general framework for building structural causal models (SCMs)
with deep learning components. The proposed approach employs normalising flows
and variational inference to enable tractable inference of exogenous noise
variables - a crucial step for counterfactual inference that is missing from
existing deep causal learning methods. Our framework is validated on a
synthetic dataset built on MNIST as well as on a real-world medical dataset of
brain MRI scans. Our experimental results indicate that we can successfully
train deep SCMs that are capable of all three levels of Pearl's ladder of
causation: association, intervention, and counterfactuals, giving rise to a
powerful new approach for answering causal questions in imaging applications
and beyond. The code for all our experiments is available at
https://github.com/biomedia-mira/deepscm.
| stat |
Estimation for network snowball sampling: Preventing pandemics | Snowball designs are the most natural of the network sampling designs. They
have many desirable properties for sampling hidden and hard-to reach
populations. They have been under-used in recent years because simple
design-based estimators and confidence intervals have not been available for
them. The needed estimation methods are supplied in this paper. Snowball
sampling methods and accurate estimators with them are needed for sampling of
the people exposed to the animals from which new coronavirus outbreaks
originate, and to sample the animal populations to which they are exposed.
Accurate estimates are needed to evaluate the effectiveness of interventions to
reduce the risk to the people exposed to the animals. In this way the
frequencies of major outbreaks and pandemics can be reduced. Snowball designs
are needed in studies of sexual and opioid networks through which HIV can
spread explosively, so that prevention intervention methods can be developed,
accurately assessed, and effectively distributed.
| stat |
Particle filters for data assimilation based on reduced order data
models | We introduce a framework for Data Assimilation (DA) in which the data is
split into multiple sets corresponding to low-rank projections of the state
space. Algorithms are developed that assimilate some or all of the projected
data, including an algorithm compatible with any generic DA method. The major
application explored here is PROJ-PF, a projected Particle Filter. The PROJ-PF
implementation assimilates highly informative but low-dimensional observations.
The implementation considered here is based upon using projections
corresponding to Assimilation in the Unstable Subspace (AUS). In the context of
particle filtering, the projected approach mitigates the collapse of particle
ensembles in high dimensional DA problems while preserving as much relevant
information as possible, as the unstable and neutral modes correspond to the
most uncertain model predictions. In particular we formulate and numerically
implement a projected Optimal Proposal Particle Filter (PROJ-OP-PF) and compare
to the standard optimal proposal and to the Ensemble Transform Kalman Filter.
| stat |
Federated Generative Privacy | In this paper, we propose FedGP, a framework for privacy-preserving data
release in the federated learning setting. We use generative adversarial
networks, generator components of which are trained by FedAvg algorithm, to
draw privacy-preserving artificial data samples and empirically assess the risk
of information disclosure. Our experiments show that FedGP is able to generate
labelled data of high quality to successfully train and validate supervised
models. Finally, we demonstrate that our approach significantly reduces
vulnerability of such models to model inversion attacks.
| stat |
Identification of Pediatric Sepsis Subphenotypes for Enhanced Machine
Learning Predictive Performance: A Latent Profile Analysis | Background: While machine learning (ML) models are rapidly emerging as
promising screening tools in critical care medicine, the identification of
homogeneous subphenotypes within populations with heterogeneous conditions such
as pediatric sepsis may facilitate attainment of high-predictive performance of
these prognostic algorithms. This study is aimed to identify subphenotypes of
pediatric sepsis and demonstrate the potential value of partitioned
data/subtyping-based training. Methods: This was a retrospective study of
clinical data extracted from medical records of 6,446 pediatric patients that
were admitted at a major hospital system in the DC area. Vitals and labs
associated with patients meeting the diagnostic criteria for sepsis were used
to perform latent profile analysis. Modern ML algorithms were used to explore
the predictive performance benefits of reduced training data heterogeneity via
label profiling. Results: In total 134 (2.1%) patients met the diagnostic
criteria for sepsis in this cohort and latent profile analysis identified four
profiles/subphenotypes of pediatric sepsis. Profiles 1 and 3 had the lowest
mortality and included pediatric patients from different age groups. Profile 2
were characterized by respiratory dysfunction; profile 4 by neurological
dysfunction and highest mortality rate (22.2%). Machine learning experiments
comparing the predictive performance of models derived without training data
profiling against profile targeted models suggest statistically significant
improved performance of prediction can be obtained. For example, area under ROC
curve (AUC) obtained to predict profile 4 with 24-hour data (AUC = .998, p <
.0001) compared favorably with the AUC obtained from the model considering all
profiles as a single homogeneous group (AUC = .918) with 24-hour data.
| stat |
Hybrid Tree-based Models for Insurance Claims | Two-part models and Tweedie generalized linear models (GLMs) have been used
to model loss costs for short-term insurance contract. For most portfolios of
insurance claims, there is typically a large proportion of zero claims that
leads to imbalances resulting in inferior prediction accuracy of these
traditional approaches. This article proposes the use of tree-based models with
a hybrid structure that involves a two-step algorithm as an alternative
approach to these traditional models. The first step is the construction of a
classification tree to build the probability model for frequency. In the second
step, we employ elastic net regression models at each terminal node from the
classification tree to build the distribution model for severity. This hybrid
structure captures the benefits of tuning hyperparameters at each step of the
algorithm; this allows for improved prediction accuracy and tuning can be
performed to meet specific business objectives. We examine and compare the
predictive performance of such a hybrid tree-based structure in relation to the
traditional Tweedie model using both real and synthetic datasets. Our empirical
results show that these hybrid tree-based models produce more accurate
predictions without the loss of intuitive interpretation.
| stat |
New algorithms and goodness-of-fit diagnostics from remarkable
properties of ranking models | The forward order assumption postulates that the ranking process of the items
is carried out by sequentially assigning the positions from the top
(most-liked) to the bottom (least-liked) alternative. This assumption has been
recently relaxed with the Extended Plackett-Luce model (EPL) through the
introduction of the discrete reference order parameter, describing the rank
attribution path. By starting from two formal properties of the EPL, the former
related to the inverse ordering of the item probabilities at the first and last
stage of the ranking process and the latter well-known as independence of
irrelevant alternatives (or Luce's choice axiom), we derive novel diagnostic
tools for testing the appropriateness of the EPL assumption as the actual
sampling distribution of the observed rankings. Besides contributing to fill
the gap of goodness-of-fit methods for the family of multistage models, we also
show how one of the two statistics can be conveniently exploited to construct a
heuristic method, that surrogates the maximum likelihood approach for inferring
the underlying reference order parameter. The relative performance of the
proposals compared with more conventional approaches is illustrated by means of
extensive simulation studies.
| stat |
Estimation of ascertainment bias and its effect on power in clinical
trials with time-to-event outcomes | While the gold standard for clinical trials is to blind all parties --
participants, researchers, and evaluators -- to treatment assignment, this is
not always a possibility. When some or all of the above individuals know the
treatment assignment, this leaves the study open to the introduction of
post-randomization biases. In the Strategies to Reduce Injuries and Develop
Confidence in Elders (STRIDE) trial, we were presented with the potential for
the unblinded clinicians administering the treatment, as well as the
individuals enrolled in the study, to introduce ascertainment bias into some
but not all events comprising the primary outcome. In this manuscript, we
present ways to estimate the ascertainment bias for a time-to-event outcome,
and discuss its impact on the overall power of a trial versus changing of the
outcome definition to a more stringent unbiased definition that restricts
attention to measurements less subject to potentially differential assessment.
We found that for the majority of situations, it is better to revise the
definition to a more stringent definition, as was done in STRIDE, even though
fewer events may be observed.
| stat |
Adversary-resilient Distributed and Decentralized Statistical Inference
and Machine Learning: An Overview of Recent Advances Under the Byzantine
Threat Model | While the last few decades have witnessed a huge body of work devoted to
inference and learning in distributed and decentralized setups, much of this
work assumes a non-adversarial setting in which individual nodes---apart from
occasional statistical failures---operate as intended within the algorithmic
framework. In recent years, however, cybersecurity threats from malicious
non-state actors and rogue entities have forced practitioners and researchers
to rethink the robustness of distributed and decentralized algorithms against
adversarial attacks. As a result, we now have a plethora of algorithmic
approaches that guarantee robustness of distributed and/or decentralized
inference and learning under different adversarial threat models. Driven in
part by the world's growing appetite for data-driven decision making, however,
securing of distributed/decentralized frameworks for inference and learning
against adversarial threats remains a rapidly evolving research area. In this
article, we provide an overview of some of the most recent developments in this
area under the threat model of Byzantine attacks.
| stat |
A cost-reducing partial labeling estimator in text classification
problem | We propose a new approach to address the text classification problems when
learning with partial labels is beneficial. Instead of offering each training
sample a set of candidate labels, we assign negative-oriented labels to the
ambiguous training examples if they are unlikely fall into certain classes. We
construct our new maximum likelihood estimators with self-correction property,
and prove that under some conditions, our estimators converge faster. Also we
discuss the advantages of applying one of our estimator to a fully supervised
learning problem. The proposed method has potential applicability in many
areas, such as crowdsourcing, natural language processing and medical image
analysis.
| stat |
Discussion of "Unbiased Markov chain Monte Carlo with couplings" by
Pierre E. Jacob, John O'Leary and Yves F. Atchad\'e | This is a contribution for the discussion on "Unbiased Markov chain Monte
Carlo with couplings" by Pierre E. Jacob, John O'Leary and Yves F. Atchad\'e to
appear in the Journal of the Royal Statistical Society Series B.
| stat |
Stopping criterion for active learning based on deterministic
generalization bounds | Active learning is a framework in which the learning machine can select the
samples to be used for training. This technique is promising, particularly when
the cost of data acquisition and labeling is high. In active learning,
determining the timing at which learning should be stopped is a critical issue.
In this study, we propose a criterion for automatically stopping active
learning. The proposed stopping criterion is based on the difference in the
expected generalization errors and hypothesis testing. We derive a novel upper
bound for the difference in expected generalization errors before and after
obtaining a new training datum based on PAC-Bayesian theory. Unlike ordinary
PAC-Bayesian bounds, though, the proposed bound is deterministic; hence, there
is no uncontrollable trade-off between the confidence and tightness of the
inequality. We combine the upper bound with a statistical test to derive a
stopping criterion for active learning. We demonstrate the effectiveness of the
proposed method via experiments with both artificial and real datasets.
| stat |
Intentional Control of Type I Error over Unconscious Data Distortion: a
Neyman-Pearson Approach to Text Classification | This paper addresses the challenges in classifying textual data obtained from
open online platforms, which are vulnerable to distortion. Most existing
classification methods minimize the overall classification error and may yield
an undesirably large type I error (relevant textual messages are classified as
irrelevant), particularly when available data exhibit an asymmetry between
relevant and irrelevant information. Data distortion exacerbates this situation
and often leads to fallacious prediction. To deal with inestimable data
distortion, we propose the use of the Neyman-Pearson (NP) classification
paradigm, which minimizes type II error under a user-specified type I error
constraint. Theoretically, we show that the NP oracle is unaffected by data
distortion when the class conditional distributions remain the same.
Empirically, we study a case of classifying posts about worker strikes obtained
from a leading Chinese microblogging platform, which are frequently prone to
extensive, unpredictable and inestimable censorship. We demonstrate that, even
though the training and test data are susceptible to different distortion and
therefore potentially follow different distributions, our proposed NP methods
control the type I error on test data at the targeted level. The methods and
implementation pipeline proposed in our case study are applicable to many other
problems involving data distortion.
| stat |
Bayesian Rank-Based Hypothesis Testing for the Rank Sum Test, the Signed
Rank Test, and Spearman's $\rho$ | Bayesian inference for rank-order problems is frustrated by the absence of an
explicit likelihood function. This hurdle can be overcome by assuming a latent
normal representation that is consistent with the ordinal information in the
data: the observed ranks are conceptualized as an impoverished reflection of an
underlying continuous scale, and inference concerns the parameters that govern
the latent representation. We apply this generic data-augmentation method to
obtain Bayes factors for three popular rank-based tests: the rank sum test, the
signed rank test, and Spearman's $\rho_s$.
| stat |
Probabilistic forecasting approaches for extreme NO$_2$ episodes: a
comparison of models | High concentration episodes for NO$_2$ are increasingly dealt with by
authorities through traffic restrictions which are activated when air quality
deteriorates beyond certain thresholds. Foreseeing the probability that
pollutant concentrations reach those thresholds becomes thus a necessity.
Probabilistic forecasting is a family of techniques that allow for the
prediction of the expected distribution function instead of a single value. In
the case of NO$_2$, it allows for the calculation of future chances of
exceeding thresholds and to detect pollution peaks. We thoroughly compared 10
state of the art probabilistic predictive models, using them to predict the
distribution of NO$_2$ concentrations in a urban location for a set of
forecasting horizons (up to 60 hours). Quantile gradient boosted trees shows
the best performance, yielding the best results for both the expected value and
the forecast full distribution. Furthermore, we show how this approach can be
used to detect pollution peaks.
| stat |
Early features associated with the neurocognitive development at 36
months of age: the AuBE study | Background. Few studies on the relations between sleep quantity and/or
quality and cognition were conducted among pre-schoolers from healthy general
population. We aimed at identifying, among 3 years old children, early factors
associated with intelligence quotient estimated through Weschler Preschool and
Primary Scale Intelligence-III test and its indicators: full-scale-, verbal-
and performance-intelligence quotients and their sub-scale scores. Methods. We
included 194 children from the French birth-cohort AuBE with both available
Weschler Preschool and Primary Scale Intelligence-III scores at 3y and sleep
data. Information was collected through self-questionnaires at birth, 6, 12, 18
and 24 months. A day/night sleep ratio was calculated. Results. Mean scores
were in normal ranges for verbal-, performance- and full-scale-intelligence
quotients. In multivariate models, being a $\ge$3 born-child and watching
television $\ge$1 hour/day at 24 months were negatively associated with all
intelligence quotient scores while collective care arrangement was positively
associated. Night waking at 6 and frequent snoring at 18 months were negatively
associated with performance intelligence quotient, some subscales and
full-scale-intelligence quotient contrary to day/night sleep ratio at 12
months. No association was observed between early sleep characteristics and
verbal intelligence quotient. Conclusion. We showed that early features
including infant sleep characteristics influence intelligence quotient scores
at 3 years old. Some of these may be accessible to prevention.
| stat |
On Structured Filtering-Clustering: Global Error Bound and Optimal
First-Order Algorithms | In recent years, the filtering-clustering problems have been a central topic
in statistics and machine learning, especially the $\ell_1$-trend filtering and
$\ell_2$-convex clustering problems. In practice, such structured problems are
typically solved by first-order algorithms despite the extremely
ill-conditioned structures of difference operator matrices. Inspired by the
desire to analyze the convergence rates of these algorithms, we show that for a
large class of filtering-clustering problems, a \textit{global error bound}
condition is satisfied for the dual filtering-clustering problems when a
certain regularization is chosen. Based on this result, we show that many
first-order algorithms attain the \textit{optimal rate of convergence} in
different settings. In particular, we establish a generalized dual gradient
ascent (GDGA) algorithmic framework with several subroutines. In deterministic
setting when the subroutine is accelerated gradient descent (AGD), the
resulting algorithm attains the linear convergence. This linear convergence
also holds for the finite-sum setting in which the subroutine is the Katyusha
algorithm. We also demonstrate that the GDGA with stochastic gradient descent
(SGD) subroutine attains the optimal rate of convergence up to the logarithmic
factor, shedding the light to the possibility of solving the
filtering-clustering problems efficiently in online setting. Experiments
conducted on $\ell_1$-trend filtering problems illustrate the favorable
performance of our algorithms over other competing algorithms.
| stat |
Hadamard Wirtinger Flow for Sparse Phase Retrieval | We consider the problem of reconstructing an $n$-dimensional $k$-sparse
signal from a set of noiseless magnitude-only measurements. Formulating the
problem as an unregularized empirical risk minimization task, we study the
sample complexity performance of gradient descent with Hadamard
parametrization, which we call Hadamard Wirtinger flow (HWF). Provided
knowledge of the signal sparsity $k$, we prove that a single step of HWF is
able to recover the support from $k(x^*_{max})^{-2}$ (modulo logarithmic term)
samples, where $x^*_{max}$ is the largest component of the signal in magnitude.
This support recovery procedure can be used to initialize existing
reconstruction methods and yields algorithms with total runtime proportional to
the cost of reading the data and improved sample complexity, which is linear in
$k$ when the signal contains at least one large component. We numerically
investigate the performance of HWF at convergence and show that, while not
requiring any explicit form of regularization nor knowledge of $k$, HWF adapts
to the signal sparsity and reconstructs sparse signals with fewer measurements
than existing gradient based methods.
| stat |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.