title
stringlengths 6
244
| abstract
stringlengths 19
5.09k
| label
stringclasses 10
values |
|---|---|---|
Random Matrix-Improved Estimation of the Wasserstein Distance between
two Centered Gaussian Distributions | This article proposes a method to consistently estimate functionals
$\frac1p\sum_{i=1}^pf(\lambda_i(C_1C_2))$ of the eigenvalues of the product of
two covariance matrices $C_1,C_2\in\mathbb{R}^{p\times p}$ based on the
empirical estimates $\lambda_i(\hat C_1\hat C_2)$ ($\hat
C_a=\frac1{n_a}\sum_{i=1}^{n_a} x_i^{(a)}x_i^{(a){{\sf T}}}$), when the size
$p$ and number $n_a$ of the (zero mean) samples $x_i^{(a)}$ are similar. As a
corollary, a consistent estimate of the Wasserstein distance (related to the
case $f(t)=\sqrt{t}$) between centered Gaussian distributions is derived.
The new estimate is shown to largely outperform the classical sample
covariance-based `plug-in' estimator. Based on this finding, a practical
application to covariance estimation is then devised which demonstrates
potentially significant performance gains with respect to state-of-the-art
alternatives.
| stat |
Improve Diverse Text Generation by Self Labeling Conditional Variational
Auto Encoder | Diversity plays a vital role in many text generating applications. In recent
years, Conditional Variational Auto Encoders (CVAE) have shown promising
performances for this task. However, they often encounter the so called
KL-Vanishing problem. Previous works mitigated such problem by heuristic
methods such as strengthening the encoder or weakening the decoder while
optimizing the CVAE objective function. Nevertheless, the optimizing direction
of these methods are implicit and it is hard to find an appropriate degree to
which these methods should be applied. In this paper, we propose an explicit
optimizing objective to complement the CVAE to directly pull away from
KL-vanishing. In fact, this objective term guides the encoder towards the "best
encoder" of the decoder to enhance the expressiveness. A labeling network is
introduced to estimate the "best encoder". It provides a continuous label in
the latent space of CVAE to help build a close connection between latent
variables and targets. The whole proposed method is named Self Labeling
CVAE~(SLCVAE). To accelerate the research of diverse text generation, we also
propose a large native one-to-many dataset. Extensive experiments are conducted
on two tasks, which show that our method largely improves the generating
diversity while achieving comparable accuracy compared with state-of-art
algorithms.
| stat |
A General Framework for Consistent Structured Prediction with Implicit
Loss Embeddings | We propose and analyze a novel theoretical and algorithmic framework for
structured prediction. While so far the term has referred to discrete output
spaces, here we consider more general settings, such as manifolds or spaces of
probability measures. We define structured prediction as a problem where the
output space lacks a vectorial structure. We identify and study a large class
of loss functions that implicitly defines a suitable geometry on the problem.
The latter is the key to develop an algorithmic framework amenable to a sharp
statistical analysis and yielding efficient computations. When dealing with
output spaces with infinite cardinality, a suitable implicit formulation of the
estimator is shown to be crucial.
| stat |
A Simple Yet Efficient Parametric Method of Local False Discovery Rate
Estimation Designed for Genome-Wide Association Data Analysis | In genome-wide association studies, hundreds of thousands of genetic features
(genes, proteins, etc.) in a given case-control population are tested to verify
existence of an association between each genetic marker and a specific disease.
A popular approach in this regard is to estimate local false discovery rate
(LFDR), the posterior probability that the null hypothesis is true, given an
observed test statistic. However, the existing LFDR estimation methods in the
literature are usually complicated. Assuming a chi-square model with one degree
of freedom, which covers many situations in genome-wide association studies, we
use the method of moments and introduce a simple, fast and efficient approach
for LFDR estimation. We perform two different simulation strategies and compare
the performance of the proposed approach with three popular LFDR estimation
methods. We also examine the practical utility of the proposed method by
analyzing a comprehensive 1000 genomes-based genome-wide association data
containing approximately 9.4 million single nucleotide polymorphisms, and a
microarray data set consisting of genetic expression levels for 6033 genes for
prostate cancer patients. The R package implementing the proposed method is
available on CRAN <https://cran.r-project.org/web/packages/LFDR.MME>.
| stat |
Direction Matters: On Influence-Preserving Graph Summarization and
Max-cut Principle for Directed Graphs | Summarizing large-scaled directed graphs into small-scale representations is
a useful but less studied problem setting. Conventional clustering approaches,
which based on "Min-Cut"-style criteria, compress both the vertices and edges
of the graph into the communities, that lead to a loss of directed edge
information. On the other hand, compressing the vertices while preserving the
directed edge information provides a way to learn the small-scale
representation of a directed graph. The reconstruction error, which measures
the edge information preserved by the summarized graph, can be used to learn
such representation. Compared to the original graphs, the summarized graphs are
easier to analyze and are capable of extracting group-level features which is
useful for efficient interventions of population behavior. In this paper, we
present a model, based on minimizing reconstruction error with non-negative
constraints, which relates to a "Max-Cut" criterion that simultaneously
identifies the compressed nodes and the directed compressed relations between
these nodes. A multiplicative update algorithm with column-wise normalization
is proposed. We further provide theoretical results on the identifiability of
the model and on the convergence of the proposed algorithms. Experiments are
conducted to demonstrate the accuracy and robustness of the proposed method.
| stat |
Monitoring and explainability of models in production | The machine learning lifecycle extends beyond the deployment stage.
Monitoring deployed models is crucial for continued provision of high quality
machine learning enabled services. Key areas include model performance and data
monitoring, detecting outliers and data drift using statistical techniques, and
providing explanations of historic predictions. We discuss the challenges to
successful implementation of solutions in each of these areas with some recent
examples of production ready solutions using open source tools.
| stat |
On default priors for robust Bayesian estimation with divergences | This paper presents objective priors for robust Bayesian estimation against
outliers based on divergences. The minimum $\gamma$-divergence estimator is
well-known to work well estimation against heavy contamination. The robust
Bayesian methods by using quasi-posterior distributions based on divergences
have been also proposed in recent years. In objective Bayesian framework, the
selection of default prior distributions under such quasi-posterior
distributions is an important problem. In this study, we provide some
properties of reference and moment matching priors under the quasi-posterior
distribution based on the $\gamma$-divergence. In particular, we show that the
proposed priors are approximately robust under the condition on the
contamination distribution without assuming any conditions on the contamination
ratio. Some simulation studies are also presented.
| stat |
A measure of evidence based on the likelihood-ratio statistics | In this paper, we show that the likelihood-ratio measure (a) is invariant
with respect to dominating sigma-finite measures, (b) satisfies logical
consequences which are not satisfied by standard $p$-values, (c) respects
frequentist properties, i.e., the type I error can be properly controlled, and,
under mild regularity conditions, (d) can be used as an upper bound for
posterior probabilities. We also discuss a generic application to test whether
the genotype frequencies of a given population are under the Hardy-Weinberg
equilibrium, under inbreeding restrictions or under outbreeding restrictions.
| stat |
Latent Gaussian process with composite likelihoods and numerical
quadrature | Clinical patient records are an example of high-dimensional data that is
typically collected from disparate sources and comprises of multiple
likelihoods with noisy as well as missing values. In this work, we propose an
unsupervised generative model that can learn a low-dimensional representation
among the observations in a latent space, while making use of all available
data in a heterogeneous data setting with missing values. We improve upon the
existing Gaussian process latent variable model (GPLVM) by incorporating
multiple likelihoods and deep neural network parameterised back-constraints to
create a non-linear dimensionality reduction technique for heterogeneous data.
In addition, we develop a variational inference method for our model that uses
numerical quadrature. We establish the effectiveness of our model and compare
against existing GPLVM methods on a standard benchmark dataset as well as on
clinical data of Parkinson's disease patients treated at the HUS Helsinki
University Hospital.
| stat |
Local polynomial regression for pooled response data | We propose local polynomial estimators for the conditional mean of a
continuous response when only pooled response data are collected under
different pooling designs. Asymptotic properties of these estimators are
investigated and compared. Extensive simulation studies are carried out to
compare finite sample performance of the proposed estimators under various
model settings and pooling strategies. We apply the proposed local polynomial
regression methods to two real-life applications to illustrate practical
implementation and performance of the estimators for the mean function.
| stat |
A machine learning approach for underwater gas leakage detection | Underwater gas reservoirs are used in many situations. In particular, Carbon
Capture and Storage (CCS) facilities that are currently being developed intend
to store greenhouse gases inside geological formations in the deep sea. In
these formations, however, the gas might percolate, leaking back to the water
and eventually to the atmosphere. The early detection of such leaks is
therefore tantamount to any underwater CCS project. In this work, we propose to
use Passive Acoustic Monitoring (PAM) and a machine learning approach to design
efficient detectors that can signal the presence of a leakage. We use data
obtained from simulation experiments off the Brazilian shore, and show that the
detection based on classification algorithms achieve good performance. We also
propose a smoothing strategy based on Hidden Markov Models in order to
incorporate previous knowledge about the probabilities of leakage occurrences.
| stat |
Model-based Inference for Rare and Clustered Populations from Adaptive
Cluster Sampling using Auxiliary Variables | Rare populations, such as endangered animals and plants, drug users and
individuals with rare diseases, tend to cluster in regions. Adaptive cluster
sampling is generally applied to obtain information from clustered and sparse
populations since it increases survey effort in areas where the individuals of
interest are observed. This work aims to propose a unit-level model which
assumes that counts are related to auxiliary variables, improving the sampling
process, assigning different weights to the cells, besides referring them
spatially. The proposed model fits rare and grouped populations, disposed over
a regular grid, in a Bayesian framework. The approach is compared to
alternative methods using simulated data and a real experiment in which
adaptive samples were drawn from an African Buffaloes population in a 24,108
square kilometers area of East Africa. Simulation studies show that the model
is efficient under several settings, validating the methodology proposed in
this paper for practical situations.
| stat |
Semi-supervised Learning of Galaxy Morphology using Equivariant
Transformer Variational Autoencoders | The growth in the number of galaxy images is much faster than the speed at
which these galaxies can be labelled by humans. However, by leveraging the
information present in the ever growing set of unlabelled images,
semi-supervised learning could be an effective way of reducing the required
labelling and increasing classification accuracy. We develop a Variational
Autoencoder (VAE) with Equivariant Transformer layers with a classifier network
from the latent space. We show that this novel architecture leads to
improvements in accuracy when used for the galaxy morphology classification
task on the Galaxy Zoo data set. In addition we show that pre-training the
classifier network as part of the VAE using the unlabelled data leads to higher
accuracy with fewer labels compared to exiting approaches. This novel VAE has
the potential to automate galaxy morphology classification with reduced human
labelling efforts.
| stat |
Permutation Inference for Canonical Correlation Analysis | Canonical correlation analysis (CCA) has become a key tool for population
neuroimaging, allowing investigation of associations between many imaging and
non-imaging measurements. As other variables are often a source of variability
not of direct interest, previous work has used CCA on residuals from a model
that removes these effects, then proceeded directly to permutation inference.
We show that such a simple permutation test leads to inflated error rates. The
reason is that residualisation introduces dependencies among the observations
that violate the exchangeability assumption. Even in the absence of nuisance
variables, however, a simple permutation test for CCA also leads to excess
error rates for all canonical correlations other than the first. The reason is
that a simple permutation scheme does not ignore the variability already
explained by previous canonical variables. Here we propose solutions for both
problems: in the case of nuisance variables, we show that transforming the
residuals to a lower dimensional basis where exchangeability holds results in a
valid permutation test; for more general cases, with or without nuisance
variables, we propose estimating the canonical correlations in a stepwise
manner, removing at each iteration the variance already explained, while
dealing with different number of variables in both sides. We also discuss how
to address the multiplicity of tests, proposing an admissible test that is not
conservative, and provide a complete algorithm for permutation inference for
CCA.
| stat |
Mobility Functional Areas and COVID-19 Spread | This work introduces a new concept of functional areas called Mobility
Functional Areas (MFAs), i.e., the geographic zones highly interconnected
according to the analysis of mobile positioning data. The MFAs do not coincide
necessarily with administrative borders as they are built observing natural
human mobility and, therefore, they can be used to inform, in a bottom-up
approach, local transportation, health and economic policies. After presenting
the methodology behind the MFAs, this study focuses on the link between the
COVID-19 pandemic and the MFAs in Austria. It emerges that the MFAs registered
an average number of infections statistically larger than the areas in the rest
of the country, suggesting the usefulness of the MFAs in the context of
targeted re-escalation policy responses to this health crisis.
| stat |
Principal Component Analysis using Frequency Components of Multivariate
Time Series | Dimension reduction techniques for multivariate time series decompose the
observed series into a few useful independent/orthogonal univariate components.
We develop a spectral domain method for multivariate second-order stationary
time series that linearly transforms the observed series into several groups of
lower-dimensional multivariate subseries. These multivariate subseries have
non-zero spectral coherence among components within a group but have zero
spectral coherence among components across groups. The observed series is
expressed as a sum of frequency components whose variances are proportional to
the spectral matrices at the respective frequencies. The demixing matrix is
then estimated using an eigendecomposition on the sum of the variance matrices
of these frequency components and its asymptotic properties are derived.
Finally, a consistent test on the cross-spectrum of pairs of components is used
to find the desired segmentation into the lower-dimensional subseries. The
numerical performance of the proposed method is illustrated through simulation
examples and an application to modeling and forecasting wind data is presented.
| stat |
The Hot Hand and Its Effect on the NBA | This paper aims to revisit and expand upon previous work on the "hot hand"
phenomenon in basketball, specifically in the NBA. Using larger, modern data
sets, we test streakiness of shooting patterns and the presence of hot hand
behavior in free throw shooting, while going further by examining league-wide
hot hand trends and the changes in individual player behavior. Additionally, we
perform simulations in order to assess their power. While we find no evidence
of the hot hand in game-play and only weak evidence in free throw trials, we
find that some NBA players exhibit behavioral changes based on the outcome of
their previous shot.
| stat |
A Model-Based General Alternative to the Standardised Precipitation
Index | In this paper, we introduce two new model-based versions of the widely-used
standardized precipitation index (SPI) for detecting and quantifying the
magnitude of extreme hydro-climatic events. Our analytical approach is based on
generalized additive models for location, scale and shape (GAMLSS), which helps
as to overcome some limitations of the SPI. We compare our model-based
standardised indices (MBSIs) with the SPI using precipitation data collected
between January 2004 - December 2013 (522 weeks) in Caapiranga, a road-less
municipality of Amazonas State. As a result, it is shown that the MBSI-1 is an
index with similar properties to the SPI, but with improved methodology. In
comparison to the SPI, our MBSI-1 index allows for the use of different
zero-augmented distributions, it works with more flexible time-scales, can be
applied to shorter records of data and also takes into account temporal
dependencies in known seasonal behaviours. Our approach is implemented in an R
package, mbsi, available from Github.
| stat |
Generalized Forward Sufficient Dimension Reduction for Categorical and
Ordinal Responses | We present a forward sufficient dimension reduction method for categorical or
ordinal responses by extending the outer product of gradients and minimum
average variance estimator to multinomial generalized linear model. Previous
work in this direction extend forward regression to binary responses, and are
applied in a pairwise manner to multinomial data, which is less efficient than
our approach. Like other forward regression-based sufficient dimension
reduction methods, our approach avoids the relatively stringent distributional
requirements necessary for inverse regression alternatives. We show consistency
of our proposed estimator and derive its convergence rate. We develop an
algorithm for our methods based on repeated applications of available
algorithms for forward regression. We also propose a clustering-based tuning
procedure to estimate the tuning parameters. The effectiveness of our estimator
and related algorithms is demonstrated via simulations and applications.
| stat |
Probabilistic elicitation of expert knowledge through assessment of
computer simulations | We present a new method for probabilistic elicitation of expert knowledge
using binary responses of human experts assessing simulated data from a
statistical model, where the parameters are subject to uncertainty. The binary
responses describe either the absolute realism of individual simulations or the
relative realism of a pair of simulations in the two alternative versions of
out approach. Each version provides a nonparametric representation of the
expert belief distribution over the values of a model parameter, without
demanding the assertion of any opinion on the parameter values themselves. Our
framework also integrates the use of active learning to efficiently query the
experts, with the possibility to additionally provide a useful misspecification
diagnostic. We validate both methods on an automatic expert judging a binomial
distribution, and on human experts judging the distribution of voters across
political parties in the United States and Norway. Both methods provide
flexible and meaningful representations of the human experts' beliefs,
correctly identifying the higher dispersion of voters between parties in
Norway.
| stat |
What is the probability that a vaccinated person is shielded from
Covid-19? A Bayesian MCMC based reanalysis of published data with emphasis on
what should be reported as `efficacy' | Based on the information communicated in press releases, and finally
published towards the end of 2020 by Pfizer, Moderna and AstraZeneca, we have
built up a simple Bayesian model, in which the main quantity of interest plays
the role of {\em vaccine efficacy} (`$\epsilon$'). The resulting Bayesian
Network is processed by a Markov Chain Monte Carlo (MCMC), implemented in JAGS
interfaced to R via rjags. As outcome, we get several probability density
functions (pdf's) of $\epsilon$, each conditioned on the data provided by the
three pharma companies. The result is rather stable against large variations of
the number of people participating in the trials and it is `somehow' in good
agreement with the results provided by the companies, in the sense that their
values correspond to the most probable value (`mode') of the pdf's resulting
from MCMC, thus reassuring us about the validity of our simple model. However
we maintain that the number to be reported as `vaccine efficacy' should be the
mean of the distribution, rather than the mode, as it was already very clear to
Laplace about 250 years ago (its `rule of succession' follows from the simplest
problem of the kind). This is particularly important in the case in which the
number of successes equals the numbers of trials, as it happens with the
efficacy against `severe forms' of infection, claimed by Moderna to be 100%.
The implication of the various uncertainties on the predicted number of
vaccinated infectees is also shown, using both MCMC and approximated formulae.
| stat |
Two Metropolis-Hastings algorithms for posterior measures with
non-Gaussian priors in infinite dimensions | We introduce two classes of Metropolis-Hastings algorithms for sampling
target measures that are absolutely continuous with respect to non-Gaussian
prior measures on infinite-dimensional Hilbert spaces. In particular, we focus
on certain classes of prior measures for which prior-reversible proposal
kernels of the autoregressive type can be designed. We then use these proposal
kernels to design algorithms that satisfy detailed balance with respect to the
target measures. Afterwards, we introduce a new class of prior measures, called
the Bessel-K priors, as a generalization of the gamma distribution to measures
in infinite dimensions. The Bessel-K priors interpolate between well-known
priors such as the gamma distribution and Besov priors and can model sparse or
compressible parameters. We present concrete instances of our algorithms for
the Bessel-K priors in the context of numerical examples in density estimation,
finite-dimensional denoising and deconvolution on the circle.
| stat |
Learning Deep Kernels for Non-Parametric Two-Sample Tests | We propose a class of kernel-based two-sample tests, which aim to determine
whether two sets of samples are drawn from the same distribution. Our tests are
constructed from kernels parameterized by deep neural nets, trained to maximize
test power. These tests adapt to variations in distribution smoothness and
shape over space, and are especially suited to high dimensions and complex
data. By contrast, the simpler kernels used in prior kernel testing work are
spatially homogeneous, and adaptive only in lengthscale. We explain how this
scheme includes popular classifier-based two-sample tests as a special case,
but improves on them in general. We provide the first proof of consistency for
the proposed adaptation method, which applies both to kernels on deep features
and to simpler radial basis kernels or multiple kernel learning. In
experiments, we establish the superior performance of our deep kernels in
hypothesis testing on benchmark and real-world data. The code of our
deep-kernel-based two sample tests is available at
https://github.com/fengliu90/DK-for-TST.
| stat |
Computational methods for Bayesian semiparametric Item Response Theory
models | Item response theory (IRT) models are widely used to obtain interpretable
inference when analyzing data from questionnaires, scaling binary responses
into continuous constructs. Typically, these models rely on a normality
assumption for the latent trait characterizing individuals in the population
under study. However, this assumption can be unrealistic and lead to biased
results. We relax the normality assumption by considering a flexible Dirichlet
Process mixture model as a nonparametric prior on the distribution of the
individual latent traits. Although this approach has been considered in the
literature before, there is a lack of comprehensive studies of such models or
general software tools. To fill this gap, we show how the NIMBLE framework for
hierarchical statistical modeling enables the use of flexible priors on the
latent trait distribution, specifically illustrating the use of Dirichlet
Process mixtures in two-parameter logistic (2PL) IRT models. We study how
different sets of constraints can lead to model identifiability and give
guidance on eliciting prior distributions. Using both simulated and real-world
data, we conduct an in-depth study of Markov chain Monte Carlo posterior
sampling efficiency for several sampling strategies. We conclude that having
access to semiparametric models can be broadly useful, as it allows inference
on the entire underlying ability distribution and its functionals, with NIMBLE
being a flexible framework for estimation of such models.
| stat |
Analysis of the Effectiveness of Face-Coverings on the Death Rate of
COVID-19 Using Machine Learning | The recent outbreak of the COVID-19 shocked humanity leading to the death of
millions of people worldwide. To stave off the spread of the virus, the
authorities in the US, employed different strategies including the mask mandate
(MM) order issued by the states' governors. Although most of the previous
studies pointed in the direction that MM can be effective in hindering the
spread of viral infections, the effectiveness of MM in reducing the degree of
exposure to the virus and, consequently, death rates remains indeterminate.
Indeed, the extent to which the degree of exposure to COVID-19 takes part in
the lethality of the virus remains unclear. In the current work, we defined a
parameter called the average death ratio as the monthly average of the ratio of
the number of daily deaths to the total number of daily cases. We utilized
survey data provided by New York Times to quantify people's abidance to the MM
order. Additionally, we implicitly addressed the extent to which people abide
by the MM order that may depend on some parameters like population, income, and
political inclination. Using different machine learning classification
algorithms we investigated how the decrease or increase in death ratio for the
counties in the US West Coast correlates with the input parameters. Our results
showed a promising score as high as 0.94 with algorithms like XGBoost, Random
Forest, and Naive Bayes. To verify the model, the best performing algorithms
were then utilized to analyze other states (Arizona, New Jersey, New York and
Texas) as test cases. The findings show an acceptable trend, further confirming
usability of the chosen features for prediction of similar cases.
| stat |
Imputation procedures in surveys using nonparametric and machine
learning methods: an empirical comparison | Nonparametric and machine learning methods are flexible methods for obtaining
accurate predictions. Nowadays, data sets with a large number of predictors and
complex structures are fairly common. In the presence of item nonresponse,
nonparametric and machine learning procedures may thus provide a useful
alternative to traditional imputation procedures for deriving a set of imputed
values. In this paper, we conduct an extensive empirical investigation that
compares a number of imputation procedures in terms of bias and efficiency in a
wide variety of settings, including high-dimensional data sets. The results
suggest that a number of machine learning procedures perform very well in terms
of bias and efficiency.
| stat |
X-Armed Bandits: Optimizing Quantiles, CVaR and Other Risks | We propose and analyze StoROO, an algorithm for risk optimization on
stochastic black-box functions derived from StoOO. Motivated by risk-averse
decision making fields like agriculture, medicine, biology or finance, we do
not focus on the mean payoff but on generic functionals of the return
distribution. We provide a generic regret analysis of StoROO and illustrate its
applicability with two examples: the optimization of quantiles and CVaR.
Inspired by the bandit literature and black-box mean optimizers, StoROO relies
on the possibility to construct confidence intervals for the targeted
functional based on random-size samples. We detail their construction in the
case of quantiles, providing tight bounds based on Kullback-Leibler divergence.
We finally present numerical experiments that show a dramatic impact of tight
bounds for the optimization of quantiles and CVaR.
| stat |
Exploiting disagreement between high-dimensional variable selectors for
uncertainty visualization | We propose Combined Selection and Uncertainty Visualizer (CSUV), which
estimates the set of true covariates in high-dimensional linear regression and
visualizes selection uncertainties by exploiting the (dis)agreement among
different base selectors. Our proposed method selects covariates that get
selected the most frequently by the different variable selection methods on
subsampled data. The method is generic and can be used with different existing
variable selection methods. We demonstrate its variable selection performance
using real and simulated data. The variable selection method and its
uncertainty illustration tool are publicly available as R package CSUV
(https://github.com/christineyuen/CSUV). The graphical tool is also available
online via https://csuv.shinyapps.io/csuv
| stat |
Sampling for Bayesian Mixture Models: MCMC with Polynomial-Time Mixing | We study the problem of sampling from the power posterior distribution in
Bayesian Gaussian mixture models, a robust version of the classical posterior.
This power posterior is known to be non-log-concave and multi-modal, which
leads to exponential mixing times for some standard MCMC algorithms. We
introduce and study the Reflected Metropolis-Hastings Random Walk (RMRW)
algorithm for sampling. For symmetric two-component Gaussian mixtures, we prove
that its mixing time is bounded as $d^{1.5}(d + \Vert \theta_{0}
\Vert^2)^{4.5}$ as long as the sample size $n$ is of the order $d (d + \Vert
\theta_{0} \Vert^2)$. Notably, this result requires no conditions on the
separation of the two means. En route to proving this bound, we establish some
new results of possible independent interest that allow for combining
Poincar\'{e} inequalities for conditional and marginal densities.
| stat |
A Universal Hierarchy of Shift-Stable Distributions and the Tradeoff
Between Stability and Performance | Many methods which find invariant predictive distributions have been
developed to learn models that can generalize to new environments without using
samples from the target distribution. However, these methods consider differing
types of shifts in environment and have been developed under disparate
frameworks, making their comparison difficult. In this paper, we provide a
unifying graphical representation of the data generating process that can
represent all such shifts. We show there is a universal hierarchy of
shift-stable distributions which correspond to operators on a graph that
disable edges. This provides the ability to compare current methods and derive
new algorithms that find optimal invariant distributions, all of which can be
mapped to the hierarchy. We theoretically and empirically show that the degree
to which stability is desirable depends on how concerned we are about large
shifts: there is a tradeoff between stability and average performance.
| stat |
High-Dimensional Uncertainty Quantification via Rank- and
Sample-Adaptive Tensor Regression | Fabrication process variations can significantly influence the performance
and yield of nano-scale electronic and photonic circuits. Stochastic spectral
methods have achieved great success in quantifying the impact of process
variations, but they suffer from the curse of dimensionality. Recently,
low-rank tensor methods have been developed to mitigate this issue, but two
fundamental challenges remain open: how to automatically determine the tensor
rank and how to adaptively pick the informative simulation samples. This paper
proposes a novel tensor regression method to address these two challenges. We
use a $\ell_{q}/ \ell_{2}$ group-sparsity regularization to determine the
tensor rank. The resulting optimization problem can be efficiently solved via
an alternating minimization solver. We also propose a two-stage adaptive
sampling method to reduce the simulation cost. Our method considers both
exploration and exploitation via the estimated Voronoi cell volume and
nonlinearity measurement respectively. The proposed model is verified with
synthetic and some realistic circuit benchmarks, on which our method can well
capture the uncertainty caused by 19 to 100 random variables with only 100 to
600 simulation samples.
| stat |
On the limitations of probabilistic claims about the probative value of
mixed DNA profile evidence | The likelihood ratio (LR) is a commonly used measure for determining the
strength of forensic match evidence. When a forensic expert determines a high
LR for DNA found at a crime scene matching the DNA profile of a suspect they
typically report that 'this provides strong support for the prosecution
hypothesis that the DNA comes from the suspect'. However, even with a high LR,
the evidence might not support the prosecution hypothesis if the defence
hypothesis used to determine the LR is not the negation of the prosecution
hypothesis (such as when the alternative is 'DNA comes from a person unrelated
to the defendant' instead of 'DNA does not come from the suspect'). For DNA
mixture profiles, especially low template DNA (LTDNA), the value of a high LR
for a 'match' - typically computed from probabilistic genotyping software - can
be especially questionable. But this is not just because of the use of
non-exhaustive hypotheses in such cases. In contrast to single profile DNA
'matches', where the only residual uncertainty is whether a person other than
the suspect has the same matching DNA profile, it is possible for all the
genotypes of the suspect's DNA profile to appear at each locus of a DNA
mixture, even though none of the contributors has that DNA profile. In fact, in
the absence of other evidence, we show it is possible to have a very high LR
for the hypothesis 'suspect is included in the mixture' even though the
posterior probability that the suspect is included is very low. Yet, in such
cases a forensic expert will generally still report a high LR as 'strong
support for the suspect being a contributor'. Our observations suggest that, in
certain circumstances, the use of the LR may have led lawyers and jurors into
grossly overestimating the probative value of a LTDNA mixed profile 'match'
| stat |
Modeling Joint Lives within Families | Family history is usually seen as a significant factor insurance companies
look at when applying for a life insurance policy. Where it is used, family
history of cardiovascular diseases, death by cancer, or family history of high
blood pressure and diabetes could result in higher premiums or no coverage at
all. In this article, we use massive (historical) data to study dependencies
between life length within families. If joint life contracts (between a husband
and a wife) have been long studied in actuarial literature, little is known
about child and parents dependencies. We illustrate those dependencies using
19th century family trees in France, and quantify implications in annuities
computations. For parents and children, we observe a modest but significant
positive association between life lengths. It yields different estimates for
remaining life expectancy, present values of annuities, or whole life insurance
guarantee, given information about the parents (such as the number of parents
alive). A similar but weaker pattern is observed when using information on
grandparents.
| stat |
PlayeRank: data-driven performance evaluation and player ranking in
soccer via a machine learning approach | The problem of evaluating the performance of soccer players is attracting the
interest of many companies and the scientific community, thanks to the
availability of massive data capturing all the events generated during a match
(e.g., tackles, passes, shots, etc.). Unfortunately, there is no consolidated
and widely accepted metric for measuring performance quality in all of its
facets. In this paper, we design and implement PlayeRank, a data-driven
framework that offers a principled multi-dimensional and role-aware evaluation
of the performance of soccer players. We build our framework by deploying a
massive dataset of soccer-logs and consisting of millions of match events
pertaining to four seasons of 18 prominent soccer competitions. By comparing
PlayeRank to known algorithms for performance evaluation in soccer, and by
exploiting a dataset of players' evaluations made by professional soccer
scouts, we show that PlayeRank significantly outperforms the competitors. We
also explore the ratings produced by {\sf PlayeRank} and discover interesting
patterns about the nature of excellent performances and what distinguishes the
top players from the others. At the end, we explore some applications of
PlayeRank -- i.e. searching players and player versatility --- showing its
flexibility and efficiency, which makes it worth to be used in the design of a
scalable platform for soccer analytics.
| stat |
Graphical Test for Discrete Uniformity and its Applications in Goodness
of Fit Evaluation and Multiple Sample Comparison | Assessing goodness of fit to a given distribution plays an important role in
computational statistics. The Probability integral transformation (PIT) can be
used to convert the question of whether a given sample originates from a
reference distribution into a problem of testing for uniformity. We present new
simulation and optimization based methods to obtain simultaneous confidence
bands for the whole empirical cumulative distribution function (ECDF) of the
PIT values under the assumption of uniformity. Simultaneous confidence bands
correspond to such confidence intervals at each point that jointly satisfy a
desired coverage. These methods can also be applied in cases where the
reference distribution is represented only by a finite sample. The confidence
bands provide an intuitive ECDF-based graphical test for uniformity, which also
provides useful information on the quality of the discrepancy. We further
extend the simulation and optimization methods to determine simultaneous
confidence bands for testing whether multiple samples come from the same
underlying distribution. This multiple sample comparison test is especially
useful in Markov chain Monte Carlo convergence diagnostics. We provide
numerical experiments to assess the properties of the tests using both
simulated and real world data and give recommendations on their practical
application in computational statistics workflows.
| stat |
A meta-algorithm for classification using random recursive tree
ensembles: A high energy physics application | The aim of this work is to propose a meta-algorithm for automatic
classification in the presence of discrete binary classes. Classifier learning
in the presence of overlapping class distributions is a challenging problem in
machine learning. Overlapping classes are described by the presence of
ambiguous areas in the feature space with a high density of points belonging to
both classes. This often occurs in real-world datasets, one such example is
numeric data denoting properties of particle decays derived from high-energy
accelerators like the Large Hadron Collider (LHC). A significant body of
research targeting the class overlap problem use ensemble classifiers to boost
the performance of algorithms by using them iteratively in multiple stages or
using multiple copies of the same model on different subsets of the input
training data. The former is called boosting and the latter is called bagging.
The algorithm proposed in this thesis targets a challenging classification
problem in high energy physics - that of improving the statistical significance
of the Higgs discovery. The underlying dataset used to train the algorithm is
experimental data built from the official ATLAS full-detector simulation with
Higgs events (signal) mixed with different background events (background) that
closely mimic the statistical properties of the signal generating class
overlap. The algorithm proposed is a variant of the classical boosted decision
tree which is known to be one of the most successful analysis techniques in
experimental physics. The algorithm utilizes a unified framework that combines
two meta-learning techniques - bagging and boosting. The results show that this
combination only works in the presence of a randomization trick in the base
learners.
| stat |
Sparse Pseudo-input Local Kriging for Large Spatial Datasets with
Exogenous Variables | We study large-scale spatial systems that contain exogenous variables, e.g.
environmental factors that are significant predictors in spatial processes.
Building predictive models for such processes is challenging because the large
numbers of observations present makes it inefficient to apply full Kriging. In
order to reduce computational complexity, this paper proposes Sparse
Pseudo-input Local Kriging (SPLK), which utilizes hyperplanes to partition a
domain into smaller subdomains and then applies a sparse approximation of the
full Kriging to each subdomain. We also develop an optimization procedure to
find the desired hyperplanes. To alleviate the problem of discontinuity in the
global predictor, we impose continuity constraints on the boundaries of the
neighboring subdomains. Furthermore, partitioning the domain into smaller
subdomains makes it possible to use different parameter values for the
covariance function in each region and, therefore, the heterogeneity in the
data structure can be effectively captured. Numerical experiments demonstrate
that SPLK outperforms, or is comparable to, the algorithms commonly applied to
spatial datasets.
| stat |
Bayesian Pseudo Posterior Mechanism under Differential Privacy | We propose a Bayesian pseudo posterior mechanism to generate record-level
synthetic datasets equipped with an $(\epsilon,\delta)-$ probabilistic
differential privacy (pDP) guarantee from any proposed synthesis model. The
pseudo posterior mechanism employs a data record-indexed, risk-based weight
vector with weights $\in [0, 1]$ to surgically downweight high-risk records for
the generation and release of record-level synthetic data. The probabilistic
differentially private pseudo posterior synthesizer constructs weights using
Lipschitz bounds under a log-pseudo likelihood utility for each data record,
which provides a practical, general formulation for using weights based on
record-level sensitivities that we show achieves dramatic improvements in the
pDP expenditure as compared to the unweighted posterior mechanism. By selecting
weights to remove likelihood contributions with non-finite log-likelihood
values, we achieve a local privacy guarantee at every sample size. We compute a
local sensitivity specific to our Consumer Expenditure Surveys dataset for
family income, published by the U.S. Bureau of Labor Statistics, and reveal
mild conditions that guarantee its asymptotic contraction to a global
sensitivity result over the space of databases under a pDP formal privacy
framework. We show that utility is better preserved for our pseudo posterior
mechanism as compared to the exponential mechanism (EM) estimated on the same
non-private synthesizer due to the use of targeted downweighting. Our results
may be applied to \emph{any} synthesizing model envisioned by the data
disseminator in a computationally tractable way that only involves estimation
of a pseudo posterior distribution for parameter(s) $\theta$, unlike recent
approaches that use naturally-bounded utility functions under application of
the EM.
| stat |
Estimating and Improving Dynamic Treatment Regimes With a Time-Varying
Instrumental Variable | Estimating dynamic treatment regimes (DTRs) from retrospective observational
data is challenging as some degree of unmeasured confounding is often expected.
In this work, we develop a framework of estimating properly defined "optimal"
DTRs with a time-varying instrumental variable (IV) when unmeasured covariates
confound the treatment and outcome, rendering the potential outcome
distributions only partially identified. We derive a novel Bellman equation
under partial identification, use it to define a generic class of estimands
(termed IV-optimal DTRs), and study the associated estimation problem. We then
extend the IV-optimality framework to tackle the policy improvement problem,
delivering IV-improved DTRs that are guaranteed to perform no worse and
potentially better than a pre-specified baseline DTR. Importantly, our
IV-improvement framework opens up the possibility of strictly improving upon
DTRs that are optimal under the no unmeasured confounding assumption (NUCA). We
demonstrate via extensive simulations the superior performance of IV-optimal
and IV-improved DTRs over the DTRs that are optimal only under the NUCA. In a
real data example, we embed retrospective observational registry data into a
natural, two-stage experiment with noncompliance using a time-varying IV and
estimate useful IV-optimal DTRs that assign mothers to high-level or low-level
neonatal intensive care units based on their prognostic variables.
| stat |
Bayesian Effect Selection in Structured Additive Distributional
Regression Models | We propose a novel spike and slab prior specification with scaled beta prime
marginals for the importance parameters of regression coefficients to allow for
general effect selection within the class of structured additive distributional
regression. This enables us to model effects on all distributional parameters
for arbitrary parametric distributions, and to consider various effect types
such as non-linear or spatial effects as well as hierarchical regression
structures. Our spike and slab prior relies on a parameter expansion that
separates blocks of regression coefficients into overall scalar importance
parameters and vectors of standardised coefficients. Hence, we can work with a
scalar quantity for effect selection instead of a possibly high-dimensional
effect vector, which yields improved shrinkage and sampling performance
compared to the classical normal-inverse-gamma prior. We investigate the
propriety of the posterior, show that the prior yields desirable shrinkage
properties, propose a way of eliciting prior parameters and provide efficient
Markov Chain Monte Carlo sampling. Using both simulated and three large-scale
data sets, we show that our approach is applicable for data with a potentially
large number of covariates, multilevel predictors accounting for hierarchically
nested data and non-standard response distributions, such as bivariate normal
or zero-inflated Poisson.
| stat |
Bayesian Parameter Identification for Jump Markov Linear Systems | This paper presents a Bayesian method for identification of jump Markov
linear system parameters. A primary motivation is to provide accurate
quantification of parameter uncertainty without relying on asymptotic in
data-length arguments. To achieve this, the paper details a particle-Gibbs
sampling approach that provides samples from the desired posterior
distribution. These samples are produced by utilising a modified discrete
particle filter and carefully chosen conjugate priors.
| stat |
Statistical Analysis of Quantum Annealing | Quantum computers use quantum resources to carry out computational tasks and
may outperform classical computers in solving certain computational problems.
Special-purpose quantum computers such as quantum annealers employ quantum
adiabatic theorem to solve combinatorial optimization problems. In this paper,
we compare classical annealings such as simulated annealing and quantum
annealings that are done by the D-Wave machines both theoretically and
numerically. We show that if the classical and quantum annealing are
characterized by equivalent Ising models, then solving an optimization problem,
i.e., finding the minimal energy of each Ising model, by the two annealing
procedures, are mathematically identical. For quantum annealing, we also derive
the probability lower-bound on successfully solving an optimization problem by
measuring the system at the end of the annealing procedure. Moreover, we
present the Markov chain Monte Carlo (MCMC) method to realize quantum annealing
by classical computers and investigate its statistical properties. In the
numerical section, we discuss the discrepancies between the MCMC based
annealing approaches and the quantum annealing approach in solving optimization
problems.
| stat |
Alternating minimization for dictionary learning: Local Convergence
Guarantees | We present theoretical guarantees for an alternating minimization algorithm
for the dictionary learning/sparse coding problem. The dictionary learning
problem is to factorize vector samples $y^{1},y^{2},\ldots, y^{n}$ into an
appropriate basis (dictionary) $A^*$ and sparse vectors $x^{1*},\ldots,x^{n*}$.
Our algorithm is a simple alternating minimization procedure that switches
between $\ell_1$ minimization and gradient descent in alternate steps.
Dictionary learning and specifically alternating minimization algorithms for
dictionary learning are well studied both theoretically and empirically.
However, in contrast to previous theoretical analyses for this problem, we
replace a condition on the operator norm (that is, the largest magnitude
singular value) of the true underlying dictionary $A^*$ with a condition on the
matrix infinity norm (that is, the largest magnitude term). Our guarantees are
under a reasonable generative model that allows for dictionaries with growing
operator norms, and can handle an arbitrary level of overcompleteness, while
having sparsity that is information theoretically optimal. We also establish
upper bounds on the sample complexity of our algorithm.
| stat |
Analysis of association football playing styles: an innovative method to
cluster networks | In this work we develop an innovative hierarchical clustering method to
divide a sample of undirected weighted networks into groups. The methodology
consists of two phases: the first phase is aimed at putting the single networks
in a broader framework by including the characteristics of the population in
the data, while the second phase creates a subdivision of the sample on the
basis of the similarity between the community structures of the processed
networks. Starting from the representation of the team's playing style as a
network, we apply the method to group the Italian Serie A teams' performances
and consequently detect the main 15 tactics shown during the 2015-2016 season.
The information obtained is used to verify the effect of the styles of play on
the number of goals scored, and we prove the key role of one of them by
implementing an extension of the Dixon and Coles model (Dixon and Coles, 1997).
| stat |
Training Variational Autoencoders with Buffered Stochastic Variational
Inference | The recognition network in deep latent variable models such as variational
autoencoders (VAEs) relies on amortized inference for efficient posterior
approximation that can scale up to large datasets. However, this technique has
also been demonstrated to select suboptimal variational parameters, often
resulting in considerable additional error called the amortization gap. To
close the amortization gap and improve the training of the generative model,
recent works have introduced an additional refinement step that applies
stochastic variational inference (SVI) to improve upon the variational
parameters returned by the amortized inference model. In this paper, we propose
the Buffered Stochastic Variational Inference (BSVI), a new refinement
procedure that makes use of SVI's sequence of intermediate variational proposal
distributions and their corresponding importance weights to construct a new
generalized importance-weighted lower bound. We demonstrate empirically that
training the variational autoencoders with BSVI consistently out-performs SVI,
yielding an improved training procedure for VAEs.
| stat |
Correlated Weights in Infinite Limits of Deep Convolutional Neural
Networks | Infinite width limits of deep neural networks often have tractable forms.
They have been used to analyse the behaviour of finite networks, as well as
being useful methods in their own right. When investigating infinitely wide
CNNs it was observed that the correlations arising from spatial weight sharing
disappear in the infinite limit. This is undesirable, as spatial correlation is
the main motivation behind CNNs. We show that the loss of this property is not
a consequence of the infinite limit, but rather of choosing an independent
weight prior. Correlating the weights maintains the correlations in the
activations. Varying the amount of correlation interpolates between
independent-weight limits and mean-pooling. Empirical evaluation of the
infinitely wide network shows that optimal performance is achieved between the
extremes, indicating that correlations can be useful.
| stat |
Variational Temporal Deep Generative Model for Radar HRRP Target
Recognition | We develop a recurrent gamma belief network (rGBN) for radar automatic target
recognition (RATR) based on high-resolution range profile (HRRP), which
characterizes the temporal dependence across the range cells of HRRP. The
proposed rGBN adopts a hierarchy of gamma distributions to build its temporal
deep generative model. For scalable training and fast out-of-sample prediction,
we propose the hybrid of a stochastic-gradient Markov chain Monte Carlo (MCMC)
and a recurrent variational inference model to perform posterior inference. To
utilize the label information to extract more discriminative latent
representations, we further propose supervised rGBN to jointly model the HRRP
samples and their corresponding labels. Experimental results on synthetic and
measured HRRP data show that the proposed models are efficient in computation,
have good classification accuracy and generalization ability, and provide
highly interpretable multi-stochastic-layer latent structure.
| stat |
Dynamic Partial Sufficient Dimension Reduction | Sufficient dimension reduction aims for reduction of dimensionality of a
regression without loss of information by replacing the original predictor with
its lower-dimensional subspace. Partial (sufficient) dimension reduction arises
when the predictors naturally fall into two sets, X and W, and we seek
dimension reduction on X alone while considering all predictors in the
regression analysis. Though partial dimension reduction is a very general
problem, only very few research results are available when W is continuous. To
the best of our knowledge, these methods generally perform poorly when X and W
are related, furthermore, none can deal with the situation where the reduced
lower-dimensional subspace of X varies dynamically with W. In this paper, We
develop a novel dynamic partial dimension reduction method, which could handle
the dynamic dimension reduction issue and also allows the dependency of X on W.
The asymptotic consistency of our method is investigated. Extensive numerical
studies and real data analysis show that our {\it Dynamic Partial Dimension
Reduction} method has superior performance comparing to the existing methods.
| stat |
Semiparametric Nonlinear Bipartite Graph Representation Learning with
Provable Guarantees | Graph representation learning is a ubiquitous task in machine learning where
the goal is to embed each vertex into a low-dimensional vector space. We
consider the bipartite graph and formalize its representation learning problem
as a statistical estimation problem of parameters in a semiparametric
exponential family distribution. The bipartite graph is assumed to be generated
by a semiparametric exponential family distribution, whose parametric component
is given by the proximity of outputs of two one-layer neural networks, while
nonparametric (nuisance) component is the base measure. Neural networks take
high-dimensional features as inputs and output embedding vectors. In this
setting, the representation learning problem is equivalent to recovering the
weight matrices. The main challenges of estimation arise from the nonlinearity
of activation functions and the nonparametric nuisance component of the
distribution. To overcome these challenges, we propose a pseudo-likelihood
objective based on the rank-order decomposition technique and focus on its
local geometry. We show that the proposed objective is strongly convex in a
neighborhood around the ground truth, so that a gradient descent-based method
achieves linear convergence rate. Moreover, we prove that the sample complexity
of the problem is linear in dimensions (up to logarithmic factors), which is
consistent with parametric Gaussian models. However, our estimator is robust to
any model misspecification within the exponential family, which is validated in
extensive experiments.
| stat |
R\'egularisation dans les Mod\`eles Lin\'eaires G\'en\'eralis\'es Mixtes
avec effet al\'eatoire autor\'egressif | We address regularised versions of the Expectation-Maximisation (EM)
algorithm for Generalised Linear Mixed Models (GLMM) in the context of panel
data (measured on several individuals at different time points). A random
response y is modelled by a GLMM, using a set X of explanatory variables and
two random effects. The first effect introduces the dependence within
individuals on which data is repeatedly collected while the second embodies the
serially correlated time-specific effect shared by all the individuals.
Variables in X are assumed many and redundant, so that regression demands
regularisation. In this context, we first propose a L2-penalised EM algorithm
for low-dimensional data, and then a supervised component-based regularised EM
algorithm for the high-dimensional case.
| stat |
Learning Large $Q$-matrix by Restricted Boltzmann Machines | Estimation of the large $Q$-matrix in Cognitive Diagnosis Models (CDMs) with
many items and latent attributes from observational data has been a huge
challenge due to its high computational cost. Borrowing ideas from deep
learning literature, we propose to learn the large $Q$-matrix by Restricted
Boltzmann Machines (RBMs) to overcome the computational difficulties. In this
paper, key relationships between RBMs and CDMs are identified. Consistent and
robust learning of the $Q$-matrix in various CDMs is shown to be valid under
certain conditions. Our simulation studies under different CDM settings show
that RBMs not only outperform the existing methods in terms of learning speed,
but also maintain good recovery accuracy of the $Q$-matrix. In the end, we
illustrate the applicability and effectiveness of our method through a real
data analysis on the Cattell's 16 personality test data set.
| stat |
Nonparametric multivariate regression estimation for circular responses | Nonparametric estimators of a regression function with circular response and
Rd-valued predictor are considered in this work. Local polynomial type
estimators are proposed and studied. Expressions for their asymptotic biases
and variances are derived, and some guidelines to select asymptotically local
optimal bandwidth matrices are also given. The finite sample behavior of the
proposed estimators is assessed through simulations and their performance is
also illustrated with a real data set.
| stat |
Negative association, ordering and convergence of resampling methods | We study convergence and convergence rates for resampling schemes. Our first
main result is a general consistency theorem based on the notion of negative
association, which is applied to establish the almost-sure weak convergence of
measures output from Kitagawa's (1996) stratified resampling method. Carpenter
et al's (1999) systematic resampling method is similar in structure but can
fail to converge depending on the order of the input samples. We introduce a
new resampling algorithm based on a stochastic rounding technique of Srinivasan
(2001), which shares some attractive properties of systematic resampling, but
which exhibits negative association and therefore converges irrespective of the
order of the input samples. We confirm a conjecture made by Kitagawa (1996)
that ordering input samples by their states in $\mathbb{R}$ yields a faster
rate of convergence; we establish that when particles are ordered using the
Hilbert curve in $\mathbb{R}^d$, the variance of the resampling error is
${\scriptscriptstyle\mathcal{O}}(N^{-(1+1/d)})$ under mild conditions, where
$N$ is the number of particles. We use these results to establish asymptotic
properties of particle algorithms based on resampling schemes that differ from
multinomial resampling.
| stat |
State Space Expectation Propagation: Efficient Inference Schemes for
Temporal Gaussian Processes | We formulate approximate Bayesian inference in non-conjugate temporal and
spatio-temporal Gaussian process models as a simple parameter update rule
applied during Kalman smoothing. This viewpoint encompasses most inference
schemes, including expectation propagation (EP), the classical (Extended,
Unscented, etc.) Kalman smoothers, and variational inference. We provide a
unifying perspective on these algorithms, showing how replacing the power EP
moment matching step with linearisation recovers the classical smoothers. EP
provides some benefits over the traditional methods via introduction of the
so-called cavity distribution, and we combine these benefits with the
computational efficiency of linearisation, providing extensive empirical
analysis demonstrating the efficacy of various algorithms under this unifying
framework. We provide a fast implementation of all methods in JAX.
| stat |
Improving upon NBA point-differential rankings | For some time, point-differential has been thought to be a better predictor
for future NBA success than pure win-loss record. Most ranking and team
performance predictions rely largely on point-differential, often with some
normalizations built-in. In this work, various capping and weighting functions
are proposed to further improve indicator performance. A gradient descent
algorithm is also employed to discover the optimized weighting/capping function
applied to individual game scores throughout the season.
| stat |
Adversarial classification: An adversarial risk analysis approach | Classification problems in security settings are usually contemplated as
confrontations in which one or more adversaries try to fool a classifier to
obtain a benefit. Most approaches to such adversarial classification problems
have focused on game theoretical ideas with strong underlying common knowledge
assumptions, which are actually not realistic in security domains. We provide
an alternative framework to such problem based on adversarial risk analysis,
which we illustrate with several examples. Computational and implementation
issues are discussed.
| stat |
The Universal Approximation Property | The universal approximation property of various machine learning models is
currently only understood on a case-by-case basis, limiting the rapid
development of new theoretically justified neural network architectures and
blurring our understanding of our current models' potential. This paper works
towards overcoming these challenges by presenting a characterization, a
representation, a construction method, and an existence result, each of which
applies to any universal approximator on most function spaces of practical
interest. Our characterization result is used to describe which activation
functions allow the feed-forward architecture to maintain its universal
approximation capabilities when multiple constraints are imposed on its final
layers and its remaining layers are only sparsely connected. These include a
rescaled and shifted Leaky ReLU activation function but not the ReLU activation
function. Our construction and representation result is used to exhibit a
simple modification of the feed-forward architecture, which can approximate any
continuous function with non-pathological growth, uniformly on the entire
Euclidean input space. This improves the known capabilities of the feed-forward
architecture.
| stat |
Multivariate Modeling for Sustainable and Resilient Infrastructure
Systems and Communities | Sustainability and resilience of urban systems are multifaceted concepts,
requiring information about multiple system attributes to adequately evaluate
and characterize. However, despite the scientific consensus on the multivariate
nature of these concepts, many of the existing techniques to model urban
sustainability and resilience are unidimensional in nature, focusing on a
characterizing a single element of highly interconnected urban systems. We
champion a paradigm shift in modeling urban sustainability and resilience,
using an integrated approach to simultaneously estimate multiple interconnected
(correlated) system attributes of sustainability and resilience as a function
of key environmental factors. We present a novel case study and review a few
recent studies to illustrate the applicability and benefits of the multivariate
approach to modeling urban sustainability and resilience. Our proposed
framework can be utilized by infrastructure managers, urban planners, and
researchers to conceptualize and assess urban sustainability and resilience
more holistically, and to better understand the key factors in advancing the
sustainability and resilience of infrastructure systems.
| stat |
Unbiased variable importance for random forests | The default variable-importance measure in random Forests, Gini importance,
has been shown to suffer from the bias of the underlying Gini-gain splitting
criterion. While the alternative permutation importance is generally accepted
as a reliable measure of variable importance, it is also computationally
demanding and suffers from other shortcomings. We propose a simple solution to
the misleading/untrustworthy Gini importance which can be viewed as an
overfitting problem: we compute the loss reduction on the out-of-bag instead of
the in-bag training samples.
| stat |
The Barker proposal: combining robustness and efficiency in
gradient-based MCMC | There is a tension between robustness and efficiency when designing Markov
chain Monte Carlo (MCMC) sampling algorithms. Here we focus on robustness with
respect to tuning parameters, showing that more sophisticated algorithms tend
to be more sensitive to the choice of step-size parameter and less robust to
heterogeneity of the distribution of interest. We characterise this phenomenon
by studying the behaviour of spectral gaps as an increasingly poor step-size is
chosen for the algorithm. Motivated by these considerations, we propose a novel
and simple gradient-based MCMC algorithm, inspired by the classical Barker
accept-reject rule, with improved robustness properties. Extensive theoretical
results, dealing with robustness to tuning, geometric ergodicity and scaling
with dimension, suggest that the novel scheme combines the robustness of simple
schemes with the efficiency of gradient-based ones. We show numerically that
this type of robustness is particularly beneficial in the context of adaptive
MCMC, giving examples where our proposed scheme significantly outperforms
state-of-the-art alternatives.
| stat |
Normalizing Flows for Probabilistic Modeling and Inference | Normalizing flows provide a general mechanism for defining expressive
probability distributions, only requiring the specification of a (usually
simple) base distribution and a series of bijective transformations. There has
been much recent work on normalizing flows, ranging from improving their
expressive power to expanding their application. We believe the field has now
matured and is in need of a unified perspective. In this review, we attempt to
provide such a perspective by describing flows through the lens of
probabilistic modeling and inference. We place special emphasis on the
fundamental principles of flow design, and discuss foundational topics such as
expressive power and computational trade-offs. We also broaden the conceptual
framing of flows by relating them to more general probability transformations.
Lastly, we summarize the use of flows for tasks such as generative modeling,
approximate inference, and supervised learning.
| stat |
Formulating causal questions and principled statistical answers | Although review papers on causal inference methods are now available, there
is a lack of introductory overviews on what they can render and on the guiding
criteria for choosing one particular method. This tutorial gives an overview in
situations where an exposure of interest is set at a chosen baseline (`point
exposure') and the target outcome arises at a later time point. We first phrase
relevant causal questions and make a case for being specific about the possible
exposure levels involved and the populations for which the question is
relevant. Using the potential outcomes framework, we describe principled
definitions of causal effects and of estimation approaches classified according
to whether they invoke the no unmeasured confounding assumption (including
outcome regression and propensity score-based methods) or an instrumental
variable with added assumptions. We discuss challenges and potential pitfalls
and illustrate application using a `simulation learner', that mimics the effect
of various breastfeeding interventions on a child's later development. This
involves a typical simulation component with generated exposure, covariate, and
outcome data that mimic those from an observational or randomised intervention
study. The simulation learner further generates various (linked) exposure types
with a set of possible values per observation unit, from which observed as well
as potential outcome data are generated. It thus provides true values of
several causal effects. R code for data generation and analysis is available on
www.ofcaus.org, where SAS and Stata code for analysis is also provided.
| stat |
Real-time forecasting of metro origin-destination matrices with
high-order weighted dynamic mode decomposition | Forecasting the short-term ridership among origin-destination pairs (OD
matrix) of a metro system is crucial in real-time metro operation. However,
this problem is notoriously difficult due to the high-dimensional, sparse,
noisy, and skewed nature of OD matrices. This paper proposes a High-order
Weighted Dynamic Mode Decomposition (HW-DMD) model for short-term metro OD
matrices forecasting. DMD uses Singular Value Decomposition (SVD) to extract
low-rank approximation from OD data, and a high-order vector autoregression
model is estimated on the reduced space for forecasting. To address a practical
issue that metro OD matrices cannot be observed in real-time, we use the
boarding demand to replace the unavailable OD matrices. Particularly, we
consider the time-evolving feature of metro systems and improve the forecast by
exponentially reducing the weights for old data. Moreover, we develop a
tailored online update algorithm for HW-DMD to update the model coefficients
daily without storing historical data or retraining. Experiments on data from a
large-scale metro system show the proposed HW-DMD is robust to the noisy and
sparse data and significantly outperforms baseline models in forecasting both
OD matrices and boarding flow. The online update algorithm also shows
consistent accuracy over a long time when maintaining an HW-DMD model at low
costs.
| stat |
Visualizing the Effects of Predictor Variables in Black Box Supervised
Learning Models | When fitting black box supervised learning models (e.g., complex trees,
neural networks, boosted trees, random forests, nearest neighbors, local
kernel-weighted methods, etc.), visualizing the main effects of the individual
predictor variables and their low-order interaction effects is often important,
and partial dependence (PD) plots are the most popular approach for
accomplishing this. However, PD plots involve a serious pitfall if the
predictor variables are far from independent, which is quite common with large
observational data sets. Namely, PD plots require extrapolation of the response
at predictor values that are far outside the multivariate envelope of the
training data, which can render the PD plots unreliable. Although marginal
plots (M plots) do not require such extrapolation, they produce substantially
biased and misleading results when the predictors are dependent, analogous to
the omitted variable bias in regression. We present a new visualization
approach that we term accumulated local effects (ALE) plots, which inherits the
desirable characteristics of PD and M plots, without inheriting their preceding
shortcomings. Like M plots, ALE plots do not require extrapolation; and like PD
plots, they are not biased by the omitted variable phenomenon. Moreover, ALE
plots are far less computationally expensive than PD plots.
| stat |
Hierarchical clustering with discrete latent variable models and the
integrated classification likelihood | Finding a set of nested partitions of a dataset is useful to uncover relevant
structure at different scales, and is often dealt with a data-dependent
methodology. In this paper, we introduce a general two-step methodology for
model-based hierarchical clustering. Considering the integrated classification
likelihood criterion as an objective function, this work applies to every
discrete latent variable models (DLVMs) where this quantity is tractable. The
first step of the methodology involves maximizing the criterion with respect to
the partition. Addressing the known problem of sub-optimal local maxima found
by greedy hill climbing heuristics, we introduce a new hybrid algorithm based
on a genetic algorithm efficiently exploring the space of solutions. The
resulting algorithm carefully combines and merges different solutions, and
allows the joint inference of the number $K$ of clusters as well as the
clusters themselves. Starting from this natural partition, the second step of
the methodology is based on a bottom-up greedy procedure to extract a hierarchy
of clusters. In a Bayesian context, this is achieved by considering the
Dirichlet cluster proportion prior parameter $\alpha$ as a regularization term
controlling the granularity of the clustering. A new approximation of the
criterion is derived as a log-linear function of $\alpha$, enabling a simple
functional form of the merge decision criterion. This second step allows the
exploration of the clustering at coarser scales. The proposed approach is
compared with existing strategies on simulated as well as real settings, and
its results are shown to be particularly relevant. A reference implementation
of this work is available in the R package greed accompanying the paper.
| stat |
Model-assisted analyses of cluster-randomized experiments | Cluster-randomized experiments are widely used due to their logistical
convenience and policy relevance. To analyze them properly, we must address the
fact that the treatment is assigned at the cluster level instead of the
individual level. Standard analytic strategies are regressions based on
individual data, cluster averages, and cluster totals, which differ when the
cluster sizes vary. These methods are often motivated by models with strong and
unverifiable assumptions, and the choice among them can be subjective. Without
any outcome modeling assumption, we evaluate these regression estimators and
the associated robust standard errors from a design-based perspective where
only the treatment assignment itself is random and controlled by the
experimenter. We demonstrate that regression based on cluster averages targets
a weighted average treatment effect, regression based on individual data is
suboptimal in terms of efficiency, and regression based on cluster totals is
consistent and more efficient with a large number of clusters. We highlight the
critical role of covariates in improving estimation efficiency, and illustrate
the efficiency gain via both simulation studies and data analysis. Moreover, we
show that the robust standard errors are convenient approximations to the true
asymptotic standard errors under the design-based perspective. Our theory holds
even when the outcome models are misspecified, so it is model-assisted rather
than model-based. We also extend the theory to a wider class of weighted
average treatment effects.
| stat |
Interpolating Predictors in High-Dimensional Factor Regression | This work studies finite-sample properties of the risk of the minimum-norm
interpolating predictor in high-dimensional regression models. If the effective
rank of the covariance matrix $\Sigma$ of the $p$ regression features is much
larger than the sample size $n$, we show that the min-norm interpolating
predictor is not desirable, as its risk approaches the risk of trivially
predicting the response by 0. However, our detailed finite-sample analysis
reveals, surprisingly, that this behavior is not present when the regression
response and the features are {\it jointly} low-dimensional, following a widely
used factor regression model. Within this popular model class, and when the
effective rank of $\Sigma$ is smaller than $n$, while still allowing for $p \gg
n$, both the bias and the variance terms of the excess risk can be controlled,
and the risk of the minimum-norm interpolating predictor approaches optimal
benchmarks. Moreover, through a detailed analysis of the bias term, we exhibit
model classes under which our upper bound on the excess risk approaches zero,
while the corresponding upper bound in the recent work arXiv:1906.11300
diverges. Furthermore, we show that the minimum-norm interpolating predictor
analyzed under the factor regression model, despite being model-agnostic and
devoid of tuning parameters, can have similar risk to predictors based on
principal components regression and ridge regression, and can improve over
LASSO based predictors, in the high-dimensional regime.
| stat |
An Online Approach to Cyberattack Detection and Localization in Smart
Grid | Complex interconnections between information technology and digital control
systems have significantly increased cybersecurity vulnerabilities in smart
grids. Cyberattacks involving data integrity can be very disruptive because of
their potential to compromise physical control by manipulating measurement
data. This is especially true in large and complex electric networks that often
rely on traditional intrusion detection systems focused on monitoring network
traffic. In this paper, we develop an online detection algorithm to detect and
localize covert attacks on smart grids. Using a network system model, we
develop a theoretical framework by characterizing a covert attack on a
generator bus in the network as sparse features in the state-estimation
residuals. We leverage such sparsity via a regularized linear regression method
to detect and localize covert attacks based on the regression coefficients. We
conduct a comprehensive numerical study on both linear and nonlinear system
models to validate our proposed method. The results show that our method
outperforms conventional methods in both detection delay and localization
accuracy.
| stat |
Healing Products of Gaussian Processes | Gaussian processes (GPs) are nonparametric Bayesian models that have been
applied to regression and classification problems. One of the approaches to
alleviate their cubic training cost is the use of local GP experts trained on
subsets of the data. In particular, product-of-expert models combine the
predictive distributions of local experts through a tractable product
operation. While these expert models allow for massively distributed
computation, their predictions typically suffer from erratic behaviour of the
mean or uncalibrated uncertainty quantification. By calibrating predictions via
a tempered softmax weighting, we provide a solution to these problems for
multiple product-of-expert models, including the generalised product of experts
and the robust Bayesian committee machine. Furthermore, we leverage the optimal
transport literature and propose a new product-of-expert model that combines
predictions of local experts by computing their Wasserstein barycenter, which
can be applied to both regression and classification.
| stat |
Kernel-Guided Training of Implicit Generative Models with Stability
Guarantees | Modern implicit generative models such as generative adversarial networks
(GANs) are generally known to suffer from issues such as instability,
uninterpretability, and difficulty in assessing their performance. If we see
these implicit models as dynamical systems, some of these issues are caused by
being unable to control their behavior in a meaningful way during the course of
training. In this work, we propose a theoretically grounded method to guide the
training trajectories of GANs by augmenting the GAN loss function with a
kernel-based regularization term that controls local and global discrepancies
between the model and true distributions. This control signal allows us to
inject prior knowledge into the model. We provide theoretical guarantees on the
stability of the resulting dynamical system and demonstrate different aspects
of it via a wide range of experiments.
| stat |
A Spectral Analysis of Dot-product Kernels | We present eigenvalue decay estimates of integral operators associated with
compositional dot-product kernels. The estimates improve on previous ones
established for power series kernels on spheres. This allows us to obtain the
volumes of balls in the corresponding reproducing kernel Hilbert spaces. We
discuss the consequences on statistical estimation with compositional dot
product kernels and highlight interesting trade-offs between the approximation
error and the statistical error depending on the number of compositions and the
smoothness of the kernels.
| stat |
Multiple Causal Inference with Latent Confounding | Causal inference from observational data requires assumptions. These
assumptions range from measuring confounders to identifying instruments.
Traditionally, causal inference assumptions have focused on estimation of
effects for a single treatment. In this work, we construct techniques for
estimation with multiple treatments in the presence of unobserved confounding.
We develop two assumptions based on shared confounding between treatments and
independence of treatments given the confounder. Together, these assumptions
lead to a confounder estimator regularized by mutual information. For this
estimator, we develop a tractable lower bound. To recover treatment effects, we
use the residual information in the treatments independent of the confounder.
We validate on simulations and an example from clinical medicine.
| stat |
Causal Effects in Twin Studies: the Role of Interference | The use of twins designs to address causal questions is becoming increasingly
popular. A standard assumption is that there is no interference between
twins---that is, no twin's exposure has a causal impact on their co-twin's
outcome. However, there may be settings in which this assumption would not
hold, and this would (1) impact the causal interpretation of parameters
obtained by commonly used existing methods; (2) change which effects are of
greatest interest; and (3) impact the conditions under which we may estimate
these effects. We explore these issues, and we derive semi-parametric efficient
estimators for causal effects in the presence of interference between twins.
Using data from the Minnesota Twin Family Study, we apply our estimators to
assess whether twins' consumption of alcohol in early adolescence may have a
causal impact on their co-twins' substance use later in life.
| stat |
Model identification for ARMA time series through convolutional neural
networks | In this paper, we use convolutional neural networks to address the problem of
model identification for autoregressive moving average time series models. We
compare the performance of several neural network architectures, trained on
simulated time series, with likelihood based methods, in particular the Akaike
and Bayesian information criteria. We find that our neural networks can
significantly outperform these likelihood based methods in terms of accuracy
and, by orders of magnitude, in terms of speed.
| stat |
Dirichlet Process Mixture Models with Shrinkage Prior | We propose Dirichlet Process Mixture (DPM) models for prediction and
cluster-wise variable selection, based on two choices of shrinkage baseline
prior distributions for the linear regression coefficients, namely the
Horseshoe prior and Normal-Gamma prior. We show in a simulation study that each
of the two proposed DPM models tend to outperform the standard DPM model based
on the non-shrinkage normal prior, in terms of predictive, variable selection,
and clustering accuracy. This is especially true for the Horseshoe model, and
when the number of covariates exceeds the within-cluster sample size. A real
data set is analyzed to illustrate the proposed modeling methodology, where
both proposed DPM models again attained better predictive accuracy.
| stat |
Bayesian Multiple Index Models for Environmental Mixtures | An important goal of environmental health research is to assess the risk
posed by mixtures of environmental exposures. Two popular classes of models for
mixtures analyses are response-surface methods and exposure-index methods.
Response-surface methods estimate high-dimensional surfaces and are thus highly
flexible but difficult to interpret. In contrast, exposure-index methods
decompose coefficients from a linear model into an overall mixture effect and
individual index weights; these models yield easily interpretable effect
estimates and efficient inferences when model assumptions hold, but, like most
parsimonious models, incur bias when these assumptions do not hold. In this
paper we propose a Bayesian multiple index model framework that combines the
strengths of each, allowing for non-linear and non-additive relationships
between exposure indices and a health outcome, while reducing the
dimensionality of the exposure vector and estimating index weights with
variable selection. This framework contains response-surface and exposure-index
models as special cases, thereby unifying the two analysis strategies. This
unification increases the range of models possible for analyzing environmental
mixtures and health, allowing one to select an appropriate analysis from a
spectrum of models varying in flexibility and interpretability. In an analysis
of the association between telomere length and 18 organic pollutants in the
National Health and Nutrition Examination Survey (NHANES), the proposed
approach fits the data as well as more complex response-surface methods and
yields more interpretable results.
| stat |
A Generalized Approach to Power Analysis for Local Average Treatment
Effects | This study introduces a new approach to power analysis in the context of
estimating a local average treatment effect (LATE), where the study subjects
exhibit noncompliance with treatment assignment. As a result of distributional
complications in the LATE context, compared to the simple ATE context, there is
currently no standard method of power analysis for the LATE. Moreover, existing
methods and commonly used substitutes - which include instrumental variable
(IV), intent-to-treat (ITT), and scaled ATE power analyses - require specifying
generally unknown variance terms and/or rely upon strong and unrealistic
assumptions, thus providing unreliable guidance on the power of tests of the
LATE. This study develops a new approach that uses standardized effect sizes to
place bounds on the power for the most commonly used estimator of the LATE, the
Wald IV estimator, whereby variance terms and distributional parameters need
not be specified nor assumed. Instead, in addition to the effect size, sample
size, and error tolerance parameters, the only other parameter that must be
specified by the researcher is the compliance rate. Additional conditions can
also be introduced to further narrow the bounds on the power calculation. The
result is a generalized approach to power analysis in the LATE context that is
simple to implement.
| stat |
Time-dependent mediators in survival analysis: Modelling direct and
indirect effects with the additive hazards model | We discuss causal mediation analyses for survival data and propose a new
approach based on the additive hazards model. The emphasis is on a dynamic
point of view, that is, understanding how the direct and indirect effects
develop over time. Hence, importantly, we allow for a time varying mediator. To
define direct and indirect effects in such a longitudinal survival setting we
take an interventional approach (Didelez (2018)) where treatment is separated
into one aspect affecting the mediator and a different aspect affecting
survival. In general, this leads to a version of the non-parametric g-formula
(Robins (1986)). In the present paper, we demonstrate that combining the
g-formula with the additive hazards model and a sequential linear model for the
mediator process results in simple and interpretable expressions for direct and
indirect effects in terms of relative survival as well as cumulative hazards.
Our results generalise and formalise the method of dynamic path analysis (Fosen
et al. (2006), Strohmaier et al. (2015)). An application to data from a
clinical trial on blood pressure medication is given.
| stat |
How to design a pre-specified statistical analysis approach to limit
p-hacking in clinical trials: the Pre-SPEC framework | Results from clinical trials can be susceptible to bias if investigators
choose their analysis approach after seeing trial data, as this can allow them
to perform multiple analyses and then choose the method that provides the most
favourable result (commonly referred to as 'p-hacking'). Pre-specification of
the planned analysis approach is essential to help reduce such bias, as it
ensures analytical methods are chosen in advance of seeing the trial data.
However, pre-specification is only effective if done in a way that does not
allow p-hacking. For example, investigators may pre-specify a certain
statistical method such as multiple imputation, but give little detail on how
it will be implemented. Because there are many different ways to perform
multiple imputation, this approach to pre-specification is ineffective, as it
still allows investigators to analyse the data in different ways before
deciding on a final approach. In this article we describe a five-point
framework (the Pre-SPEC framework) for designing a pre-specified analysis
approach that does not allow p-hacking. This framework is intended to be used
in conjunction with the SPIRIT (Standard Protocol Items: Recommendations for
Interventional Trials) statement and other similar guidelines to help
investigators design the statistical analysis strategy for the trial's primary
outcome in the trial protocol.
| stat |
Improving linear quantile regression for replicated data | This paper deals with improvement of linear quantile regression, when there
are a few distinct values of the covariates but many replicates. On can improve
asymptotic efficiency of the estimated regression coefficients by using
suitable weights in quantile regression, or simply by using weighted least
squares regression on the conditional sample quantiles. The asymptotic
variances of the unweighted and weighted estimators coincide only in some
restrictive special cases, e.g., when the density of the conditional response
has identical values at the quantile of interest over the support of the
covariate. The dominance of the weighted estimators is demonstrated in a
simulation study, and through the analysis of a data set on tropical cyclones.
| stat |
Meta Learning for Causal Direction | The inaccessibility of controlled randomized trials due to inherent
constraints in many fields of science has been a fundamental issue in causal
inference. In this paper, we focus on distinguishing the cause from effect in
the bivariate setting under limited observational data. Based on recent
developments in meta learning as well as in causal inference, we introduce a
novel generative model that allows distinguishing cause and effect in the small
data setting. Using a learnt task variable that contains distributional
information of each dataset, we propose an end-to-end algorithm that makes use
of similar training datasets at test time. We demonstrate our method on various
synthetic as well as real-world data and show that it is able to maintain high
accuracy in detecting directions across varying dataset sizes.
| stat |
On the Effect of Suboptimal Estimation of Mutual Information in Feature
Selection and Classification | This paper introduces a new property of estimators of the strength of
statistical association, which helps characterize how well an estimator will
perform in scenarios where dependencies between continuous and discrete random
variables need to be rank ordered. The new property, termed the estimator
response curve, is easily computable and provides a marginal distribution
agnostic way to assess an estimator's performance. It overcomes notable
drawbacks of current metrics of assessment, including statistical power, bias,
and consistency. We utilize the estimator response curve to test various
measures of the strength of association that satisfy the data processing
inequality (DPI), and show that the CIM estimator's performance compares
favorably to kNN, vME, AP, and H_{MI} estimators of mutual information. The
estimators which were identified to be suboptimal, according to the estimator
response curve, perform worse than the more optimal estimators when tested with
real-world data from four different areas of science, all with varying
dimensionalities and sizes.
| stat |
Sticky PDMP samplers for sparse and local inference problems | We construct a new class of efficient Monte Carlo methods based on
continuous-time piecewise deterministic Markov processes (PDMP) suitable for
inference in high dimensional sparse models, i.e. models for which there is
prior knowledge that many coordinates are likely to be exactly $0$. This is
achieved with the fairly simple idea of endowing existing PDMP samplers with
sticky coordinate axes, coordinate planes etc. Upon hitting those subspaces, an
event is triggered, during which the process sticks to the subspace, this way
spending some time in a sub-model. That introduces non-reversible jumps between
different (sub-)models. The approach can also be combined with local
implementations of PDMP samplers to target measures that additionally exhibit a
sparse dependency structure. We illustrate the new method for a number of
statistical models where both the sample size $N$ and the dimensionality $d$ of
the parameter space are large.
| stat |
Meta-Learning surrogate models for sequential decision making | We introduce a unified probabilistic framework for solving sequential
decision making problems ranging from Bayesian optimisation to contextual
bandits and reinforcement learning. This is accomplished by a probabilistic
model-based approach that explains observed data while capturing predictive
uncertainty during the decision making process. Crucially, this probabilistic
model is chosen to be a Meta-Learning system that allows learning from a
distribution of related problems, allowing data efficient adaptation to a
target task. As a suitable instantiation of this framework, we explore the use
of Neural processes due to statistical and computational desiderata. We apply
our framework to a broad range of problem domains, such as control problems,
recommender systems and adversarial attacks on RL agents, demonstrating an
efficient and general black-box learning approach.
| stat |
NeuTra-lizing Bad Geometry in Hamiltonian Monte Carlo Using Neural
Transport | Hamiltonian Monte Carlo is a powerful algorithm for sampling from
difficult-to-normalize posterior distributions. However, when the geometry of
the posterior is unfavorable, it may take many expensive evaluations of the
target distribution and its gradient to converge and mix. We propose neural
transport (NeuTra) HMC, a technique for learning to correct this sort of
unfavorable geometry using inverse autoregressive flows (IAF), a powerful
neural variational inference technique. The IAF is trained to minimize the KL
divergence from an isotropic Gaussian to the warped posterior, and then HMC
sampling is performed in the warped space. We evaluate NeuTra HMC on a variety
of synthetic and real problems, and find that it significantly outperforms
vanilla HMC both in time to reach the stationary distribution and asymptotic
effective-sample-size rates.
| stat |
Semi-Supervised Regression using Cluster Ensemble and Low-Rank
Co-Association Matrix Decomposition under Uncertainties | In this paper, we solve a semi-supervised regression problem. Due to the lack
of knowledge about the data structure and the presence of random noise, the
considered data model is uncertain. We propose a method which combines graph
Laplacian regularization and cluster ensemble methodologies. The co-association
matrix of the ensemble is calculated on both labeled and unlabeled data; this
matrix is used as a similarity matrix in the regularization framework to derive
the predicted outputs. We use the low-rank decomposition of the co-association
matrix to significantly speedup calculations and reduce memory. Numerical
experiments using the Monte Carlo approach demonstrate robustness, efficiency,
and scalability of the proposed method.
| stat |
Linear-time Learning on Distributions with Approximate Kernel Embeddings | Many interesting machine learning problems are best posed by considering
instances that are distributions, or sample sets drawn from distributions.
Previous work devoted to machine learning tasks with distributional inputs has
done so through pairwise kernel evaluations between pdfs (or sample sets).
While such an approach is fine for smaller datasets, the computation of an $N
\times N$ Gram matrix is prohibitive in large datasets. Recent scalable
estimators that work over pdfs have done so only with kernels that use
Euclidean metrics, like the $L_2$ distance. However, there are a myriad of
other useful metrics available, such as total variation, Hellinger distance,
and the Jensen-Shannon divergence. This work develops the first random features
for pdfs whose dot product approximates kernels using these non-Euclidean
metrics, allowing estimators using such kernels to scale to large datasets by
working in a primal space, without computing large Gram matrices. We provide an
analysis of the approximation error in using our proposed random features and
show empirically the quality of our approximation both in estimating a Gram
matrix and in solving learning tasks in real-world and synthetic data.
| stat |
Learning While Dissipating Information: Understanding the Generalization
Capability of SGLD | Understanding the generalization capability of learning algorithms is at the
heart of statistical learning theory. In this paper, we investigate the
generalization gap of stochastic gradient Langevin dynamics (SGLD), a widely
used optimizer for training deep neural networks (DNNs). We derive an
algorithm-dependent generalization bound by analyzing SGLD through an
information-theoretic lens. Our analysis reveals an intricate trade-off between
learning and information dissipation: SGLD learns from data by updating
parameters at each iteration while dissipating information from early training
stages. Our bound also involves the variance of gradients which captures a
particular kind of "sharpness" of the loss landscape. The main proof techniques
in this paper rely on strong data processing inequalities -- a fundamental
concept in information theory -- and Otto-Villani's HWI inequality. Finally, we
demonstrate our bound through numerical experiments, showing that it can
predict the behavior of the true generalization gap.
| stat |
Integrative Bayesian models using Post-selective Inference: a case study
in Radiogenomics | Identifying direct links between genomic pathways and clinical endpoints for
highly fatal diseases such as cancer is a formidable task. By selecting
statistically relevant associations between a wealth of intermediary variables
such as imaging and genomic measurements, integrative analyses can potentially
result in sharper clinical models with interpretable parameters, in terms of
their mechanisms. Estimates of uncertainty in the resulting models are however
unreliable unless inference accounts for the preceding steps of selection. In
this article, we develop selection-aware Bayesian methods which are: (i)
amenable to a flexible class of integrative Bayesian models post a selection of
promising variables via $\ell_1$-regularized algorithms; (ii) enjoy
computational efficiency due to a focus on sharp models with meaning; (iii)
strike a crucial tradeoff between the quality of model selection and
inferential power. Central to our selection-aware workflow, a conditional
likelihood constructed with a reparameterization map is deployed for obtaining
uncertainty estimates in integrative models. Investigating the potential of our
methods in a radiogenomic analysis, we successfully recover several important
gene pathways and calibrate uncertainties for their associations with patient
survival times.
| stat |
Optimal Scaling of Random-Walk Metropolis Algorithms on General Target
Distributions | One main limitation of the existing optimal scaling results for
Metropolis--Hastings algorithms is that the assumptions on the target
distribution are unrealistic. In this paper, we consider optimal scaling of
random-walk Metropolis algorithms on general target distributions in high
dimensions arising from practical MCMC models from Bayesian statistics. For
optimal scaling by maximizing expected squared jumping distance (ESJD), we show
the asymptotically optimal acceptance rate $0.234$ can be obtained under
general realistic sufficient conditions on the target distribution. The new
sufficient conditions are easy to be verified and may hold for some general
classes of MCMC models arising from Bayesian statistics applications, which
substantially generalize the product i.i.d. condition required in most existing
literature of optimal scaling. Furthermore, we show one-dimensional diffusion
limits can be obtained under slightly stronger conditions, which still allow
dependent coordinates of the target distribution. We also connect the new
diffusion limit results to complexity bounds of Metropolis algorithms in high
dimensions.
| stat |
A partial least squares approach for function-on-function interaction
regression | A partial least squares regression is proposed for estimating the
function-on-function regression model where a functional response and multiple
functional predictors consist of random curves with quadratic and interaction
effects. The direct estimation of a function-on-function regression model is
usually an ill-posed problem. To overcome this difficulty, in practice, the
functional data that belong to the infinite-dimensional space are generally
projected into a finite-dimensional space of basis functions. The
function-on-function regression model is converted to a multivariate regression
model of the basis expansion coefficients. In the estimation phase of the
proposed method, the functional variables are approximated by a
finite-dimensional basis function expansion method. We show that the partial
least squares regression constructed via a functional response, multiple
functional predictors, and quadratic/interaction terms of the functional
predictors is equivalent to the partial least squares regression constructed
using basis expansions of functional variables. From the partial least squares
regression of the basis expansions of functional variables, we provide an
explicit formula for the partial least squares estimate of the coefficient
function of the function-on-function regression model. Because the true forms
of the models are generally unspecified, we propose a forward procedure for
model selection. The finite sample performance of the proposed method is
examined using several Monte Carlo experiments and two empirical data analyses,
and the results were found to compare favorably with an existing method.
| stat |
Evaluating the Implicit Midpoint Integrator for Riemannian Manifold
Hamiltonian Monte Carlo | Riemannian manifold Hamiltonian Monte Carlo is traditionally carried out
using the generalized leapfrog integrator. However, this integrator is not the
only choice and other integrators yielding valid Markov chain transition
operators may be considered. In this work, we examine the implicit midpoint
integrator as an alternative to the generalized leapfrog integrator. We discuss
advantages and disadvantages of the implicit midpoint integrator for
Hamiltonian Monte Carlo, its theoretical properties, and an empirical
assessment of the critical attributes of such an integrator for Hamiltonian
Monte Carlo: energy conservation, volume preservation, and reversibility.
Empirically, we find that while leapfrog iterations are faster, the implicit
midpoint integrator has better energy conservation, leading to higher
acceptance rates, as well as better conservation of volume and better
reversibility, arguably yielding a more accurate sampling procedure.
| stat |
Variational Nonlinear State Estimation | In this paper, the problem of state estimation, in the context of both
filtering and smoothing, for nonlinear state-space models is considered. Due to
the nonlinear nature of the models, the state estimation problem is generally
intractable as it involves integrals of general nonlinear functions and
probability density functions lacking closed-form solutions. As such, it is
common to approximate the state estimation problem. In this paper, we develop a
solution based on the variational inference principle, which offers the key
advantage of a flexible, but principled, mechanism for approximating the
required distributions. Our main contribution lies in a new formulation of the
state estimation problem as an optimisation problem, which can then be solved
using standard optimisation routines that employ exact first- and second-order
derivatives. The resulting state estimation approach involves a minimal number
of assumptions and is directly applicable to nonlinear systems with both
Gaussian and non-Gaussian probabilistic models. The performance of our approach
is demonstrated on several examples; a challenging scalar system, a model of a
simple robotic system, and a target tracking problem using a von Mises-Fisher
distribution and outperforms alternative assumed density approaches to state
estimation.
| stat |
Global Convergence of Langevin Dynamics Based Algorithms for Nonconvex
Optimization | We present a unified framework to analyze the global convergence of Langevin
dynamics based algorithms for nonconvex finite-sum optimization with $n$
component functions. At the core of our analysis is a direct analysis of the
ergodicity of the numerical approximations to Langevin dynamics, which leads to
faster convergence rates. Specifically, we show that gradient Langevin dynamics
(GLD) and stochastic gradient Langevin dynamics (SGLD) converge to the almost
minimizer within $\tilde O\big(nd/(\lambda\epsilon) \big)$ and $\tilde
O\big(d^7/(\lambda^5\epsilon^5) \big)$ stochastic gradient evaluations
respectively, where $d$ is the problem dimension, and $\lambda$ is the spectral
gap of the Markov chain generated by GLD. Both results improve upon the best
known gradient complexity results (Raginsky et al., 2017). Furthermore, for the
first time we prove the global convergence guarantee for variance reduced
stochastic gradient Langevin dynamics (SVRG-LD) to the almost minimizer within
$\tilde O\big(\sqrt{n}d^5/(\lambda^4\epsilon^{5/2})\big)$ stochastic gradient
evaluations, which outperforms the gradient complexities of GLD and SGLD in a
wide regime. Our theoretical analyses shed some light on using Langevin
dynamics based algorithms for nonconvex optimization with provable guarantees.
| stat |
Geometric Wavelet Scattering Networks on Compact Riemannian Manifolds | The Euclidean scattering transform was introduced nearly a decade ago to
improve the mathematical understanding of convolutional neural networks.
Inspired by recent interest in geometric deep learning, which aims to
generalize convolutional neural networks to manifold and graph-structured
domains, we define a geometric scattering transform on manifolds. Similar to
the Euclidean scattering transform, the geometric scattering transform is based
on a cascade of wavelet filters and pointwise nonlinearities. It is invariant
to local isometries and stable to certain types of diffeomorphisms. Empirical
results demonstrate its utility on several geometric learning tasks. Our
results generalize the deformation stability and local translation invariance
of Euclidean scattering, and demonstrate the importance of linking the used
filter structures to the underlying geometry of the data.
| stat |
Independent Action Models and Prediction of Combination Treatment
Effects for Response Rate, Duration of Response and Tumor Size Change in
Oncology Drug Development | An unprecedented number of new cancer targets are in development, and most
are being developed in combination therapies. Early oncology development is
strategically challenged in choosing the best combinations to move forward to
late stage development. The most common early endpoints to be assessed in such
decision-making include objective response rate, duration of response and tumor
size change. In this paper, using independent-drug-action and
Bliss-drug-independence concepts as a foundation, we introduce simple models to
predict combination therapy efficacy for duration of response and tumor size
change. These models complement previous publications using the independent
action models (Palmer 2017, Schmidt 2020) to predict progression-free survival
and objective response rate and serve as new predictive models to understand
drug combinations for early endpoints. The models can be applied to predict the
combination treatment effect for early endpoints given monotherapy data, or to
estimate the possible effect of one monotherapy in the combination if data are
available from the combination therapy and the other monotherapy. Such
quantitative work facilitates efficient oncology drug development.
| stat |
Bayesian Variable Selection for Cox Regression Model with Spatially
Varying Coefficients with Applications to Louisiana Respiratory Cancer Data | The Cox regression model is a commonly used model in survival analysis. In
public health studies, clinical data are often collected from medical service
providers of different locations. There are large geographical variations in
the covariate effects on survival rates from particular diseases. In this
paper, we focus on the variable selection issue for the Cox regression model
with spatially varying coefficients. We propose a Bayesian hierarchical model
which incorporates a horseshoe prior for sparsity and a point mass mixture
prior to determine whether a regression coefficient is spatially varying. An
efficient two-stage computational method is used for posterior inference and
variable selection. It essentially applies the existing method for maximizing
the partial likelihood for the Cox model by site independently first, and then
applying an MCMC algorithm for variable selection based on results of the first
stage. Extensive simulation studies are carried out to examine the empirical
performance of the proposed method. Finally, we apply the proposed methodology
to analyzing a real data set on respiratory cancer in Louisiana from the SEER
program.
| stat |
Learning Rates as a Function of Batch Size: A Random Matrix Theory
Approach to Neural Network Training | We study the effect of mini-batching on the loss landscape of deep neural
networks using spiked, field-dependent random matrix theory. We demonstrate
that the magnitude of the extremal values of the batch Hessian are larger than
those of the empirical Hessian. We also derive similar results for the
Generalised Gauss-Newton matrix approximation of the Hessian. As a consequence
of our theorems we derive an analytical expressions for the maximal learning
rates as a function of batch size, informing practical optimisation schemes for
both stochastic gradient descent (linear scaling) and adaptive algorithms such
as Adam (square root scaling). Whilst the linear scaling for stochastic
gradient descent has been derived under more restrictive conditions, which we
generalise, the square root scaling rule for adaptive optimisers is, to our
knowledge, completely novel. We validate our claims on the VGG/WideResNet
architectures on the CIFAR-100 and ImageNet datasets.
| stat |
Copula Modeling of Multivariate Longitudinal Data with Dropout | Joint multivariate longitudinal and time-to-event data are gaining increasing
attention in the biomedical sciences where subjects are followed over time to
monitor the progress of a disease or medical condition. In the insurance
context, claims outcomes may be related to a policyholder's dropout or decision
to lapse a policy. This paper introduces a generalized method of moments
technique to estimate dependence parameters where associations are represented
using copulas. A simulation study demonstrates the viability of the approach.
The paper describes how the joint model provides new information that insurers
can use to better manage their portfolios of risks using illustrative data from
a Spanish insurer.
| stat |
Wavelet-Based Moment-Matching Techniques for Inertial Sensor Calibration | The task of inertial sensor calibration has required the development of
various techniques to take into account the sources of measurement error coming
from such devices. The calibration of the stochastic errors of these sensors
has been the focus of increasing amount of research in which the method of
reference has been the so-called "Allan variance slope method" which, in
addition to not having appropriate statistical properties, requires a
subjective input which makes it prone to mistakes. To overcome this, recent
research has started proposing "automatic" approaches where the parameters of
the probabilistic models underlying the error signals are estimated by matching
functions of the Allan variance or Wavelet Variance with their model-implied
counterparts. However, given the increased use of such techniques, there has
been no study or clear direction for practitioners on which approach is optimal
for the purpose of sensor calibration. This paper formally defines the class of
estimators based on this technique and puts forward theoretical and applied
results that, comparing with estimators in this class, suggest the use of the
Generalized Method of Wavelet Moments as an optimal choice.
| stat |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.