title
stringlengths 6
244
| abstract
stringlengths 19
5.09k
| label
stringclasses 10
values |
|---|---|---|
R friendly multi-threading in C++ | Calling multi-threaded C++ code from R has its perils. Since the R
interpreter is single-threaded, one must not check for user interruptions or
print to the R console from multiple threads. One can, however, synchronize
with R from the main thread. The R package RcppThread (current version 1.0.0)
contains a header only C++ library for thread safe communication with R that
exploits this fact. It includes C++ classes for threads, a thread pool, and
parallel loops that routinely synchronize with R. This article explains the
package's functionality and gives examples of its usage. The synchronization
mechanism may also apply to other threading frameworks. Benchmarks suggest
that, although synchronization causes overhead, the parallel abstractions of
RcppThread are competitive with other popular libraries in typical scenarios
encountered in statistical computing.
| stat |
Joint Community Detection and Rotational Synchronization via
Semidefinite Programming | In the presence of heterogeneous data, where randomly rotated objects fall
into multiple underlying categories, it is challenging to simultaneously
classify them into clusters and synchronize them based on pairwise relations.
This gives rise to the joint problem of community detection and
synchronization. We propose a series of semidefinite relaxations, and prove
their exact recovery when extending the celebrated stochastic block model to
this new setting where both rotations and cluster identities are to be
determined. Numerical experiments demonstrate the efficacy of our proposed
algorithms and confirm our theoretical result which indicates a sharp phase
transition for exact recovery.
| stat |
Efficient Particle Smoothing for Bayesian Inference in Dynamic Survival
Models | This article proposes an efficient Bayesian inference for piecewise
exponential hazard (PEH) models, which allow the effect of a covariate on the
survival time to vary over time. The proposed inference methodology is based on
a particle smoothing (PS) algorithm that depends on three particle filters.
Efficient proposal (importance) distributions for the particle filters tailored
to the nature of survival data and PEH models are developed using the Laplace
approximation of the posterior distribution and linear Bayes theory. The
algorithm is applied to both simulated and real data, and the results show that
it generates an effective sample size that is more than two orders of magnitude
larger than a state-of-the-art MCMC sampler for the same computing time, and
scales well in high-dimensional and relatively large data.
| stat |
Multicategory Angle-based Learning for Estimating Optimal Dynamic
Treatment Regimes with Censored Data | An optimal dynamic treatment regime (DTR) consists of a sequence of decision
rules in maximizing long-term benefits, which is applicable for chronic
diseases such as HIV infection or cancer. In this paper, we develop a novel
angle-based approach to search the optimal DTR under a multicategory treatment
framework for survival data. The proposed method targets maximization the
conditional survival function of patients following a DTR. In contrast to most
existing approaches which are designed to maximize the expected survival time
under a binary treatment framework, the proposed method solves the
multicategory treatment problem given multiple stages for censored data.
Specifically, the proposed method obtains the optimal DTR via integrating
estimations of decision rules at multiple stages into a single multicategory
classification algorithm without imposing additional constraints, which is also
more computationally efficient and robust. In theory, we establish Fisher
consistency of the proposed method under regularity conditions. Our numerical
studies show that the proposed method outperforms competing methods in terms of
maximizing the conditional survival function. We apply the proposed method to
two real datasets: Framingham heart study data and acquired immunodeficiency
syndrome (AIDS) clinical data.
| stat |
Attention-based Multi-Input Deep Learning Architecture for Biological
Activity Prediction: An Application in EGFR Inhibitors | Machine learning and deep learning have gained popularity and achieved
immense success in Drug discovery in recent decades. Historically, machine
learning and deep learning models were trained on either structural data or
chemical properties by separated model. In this study, we proposed an
architecture training simultaneously both type of data in order to improve the
overall performance. Given the molecular structure in the form of SMILES
notation and their label, we generated the SMILES-based feature matrix and
molecular descriptors. These data were trained on a deep learning model which
was also integrated with the Attention mechanism to facilitate training and
interpreting. Experiments showed that our model could raise the performance of
prediction comparing to the reference. With the maximum MCC 0.58 and AUC 90% by
cross-validation on EGFR inhibitors dataset, our architecture was outperforming
the referring model. We also successfully integrated Attention mechanism into
our model, which helped to interpret the contribution of chemical structures on
bioactivity.
| stat |
Designed Quadrature to Approximate Integrals in Maximum Simulated
Likelihood Estimation | Maximum simulated likelihood estimation of mixed multinomial logit (MMNL) or
probit models requires evaluation of a multidimensional integral. Quasi-Monte
Carlo (QMC) methods such as shuffled and scrambled Halton sequences and
modified Latin hypercube sampling (MLHS) are workhorse methods for integral
approximation. A few earlier studies explored the potential of sparse grid
quadrature (SGQ), but this approximation suffers from negative weights. As an
alternative to QMC and SGQ, we looked into the recently developed designed
quadrature (DQ) method. DQ requires fewer nodes to get the same level of
accuracy as of QMC and SGQ, is as easy to implement, ensures positivity of
weights, and can be created on any general polynomial spaces. We benchmarked DQ
against QMC in a Monte Carlo study under different data generating processes
with a varying number of random parameters (3, 5, and 10) and
variance-covariance structures (diagonal and full). Whereas DQ significantly
outperformed QMC in the diagonal variance-covariance scenario, it could also
achieve a better model fit and recover true parameters with fewer nodes (i.e.,
relatively lower computation time) in the full variance-covariance scenario.
Finally, we evaluated the performance of DQ in a case study to understand
preferences for mobility-on-demand services in New York City. In estimating
MMNL with five random parameters, DQ achieved better fit and statistical
significance of parameters with just 200 nodes as compared to 1000 QMC draws,
making DQ around five times faster than QMC methods.
| stat |
Towards Deep Learning Models Resistant to Large Perturbations | Adversarial robustness has proven to be a required property of machine
learning algorithms. A key and often overlooked aspect of this problem is to
try to make the adversarial noise magnitude as large as possible to enhance the
benefits of the model robustness. We show that the well-established algorithm
called "adversarial training" fails to train a deep neural network given a
large, but reasonable, perturbation magnitude. In this paper, we propose a
simple yet effective initialization of the network weights that makes learning
on higher levels of noise possible. We next evaluate this idea rigorously on
MNIST ($\epsilon$ up to $\approx 0.40$) and CIFAR10 ($\epsilon$ up to $\approx
32/255$) datasets assuming the $\ell_{\infty}$ attack model. Additionally, in
order to establish the limits of $\epsilon$ in which the learning is feasible,
we study the optimal robust classifier assuming full access to the joint data
and label distribution. Then, we provide some theoretical results on the
adversarial accuracy for a simple multi-dimensional Bernoulli distribution,
which yields some insights on the range of feasible perturbations for the MNIST
dataset.
| stat |
Evolutionary Variational Optimization of Generative Models | We combine two popular optimization approaches to derive learning algorithms
for generative models: variational optimization and evolutionary algorithms.
The combination is realized for generative models with discrete latents by
using truncated posteriors as the family of variational distributions. The
variational parameters of truncated posteriors are sets of latent states. By
interpreting these states as genomes of individuals and by using the
variational lower bound to define a fitness, we can apply evolutionary
algorithms to realize the variational loop. The used variational distributions
are very flexible and we show that evolutionary algorithms can effectively and
efficiently optimize the variational bound. Furthermore, the variational loop
is generally applicable ("black box") with no analytical derivations required.
To show general applicability, we apply the approach to three generative models
(we use noisy-OR Bayes Nets, Binary Sparse Coding, and Spike-and-Slab Sparse
Coding). To demonstrate effectiveness and efficiency of the novel variational
approach, we use the standard competitive benchmarks of image denoising and
inpainting. The benchmarks allow quantitative comparisons to a wide range of
methods including probabilistic approaches, deep deterministic and generative
networks, and non-local image processing methods. In the category of
"zero-shot" learning (when only the corrupted image is used for training), we
observed the evolutionary variational algorithm to significantly improve the
state-of-the-art in many benchmark settings. For one well-known inpainting
benchmark, we also observed state-of-the-art performance across all categories
of algorithms although we only train on the corrupted image. In general, our
investigations highlight the importance of research on optimization methods for
generative models to achieve performance improvements.
| stat |
A Bayesian Framework for Generation of Fully Synthetic Mixed Datasets | Much of the micro data used for epidemiological studies contain sensitive
measurements on real individuals. As a result, such micro data cannot be
published out of privacy concerns, rendering any published statistical analyses
on them nearly impossible to reproduce. To promote the dissemination of key
datasets for analysis without jeopardizing the privacy of individuals, we
introduce a cohesive Bayesian framework for the generation of fully synthetic,
high dimensional micro datasets of mixed categorical, binary, count, and
continuous variables. This process centers around a joint Bayesian model that
is simultaneously compatible with all of these data types, enabling the
creation of mixed synthetic datasets through posterior predictive sampling.
Furthermore, a focal point of epidemiological data analysis is the study of
conditional relationships between various exposures and key outcome variables
through regression analysis. We design a modified data synthesis strategy to
target and preserve these conditional relationships, including both
nonlinearities and interactions. The proposed techniques are deployed to create
a synthetic version of a confidential dataset containing dozens of health,
cognitive, and social measurements on nearly 20,000 North Carolina children.
| stat |
The Ci3+3 Design for Dual-Agent Combination Dose-Finding Clinical Trials | We propose a rule-based statistical design for combination dose-finding
trials with two agents. The Ci3+3 design is an extension of the i3+3 design
with simple decision rules comparing the observed toxicity rates and
equivalence intervals that define the maximum tolerated dose combination. Ci3+3
consists of two stages to allow fast and efficient exploration of the
dose-combination space. Statistical inference is restricted to a beta-binomial
model for dose evaluation, and the entire design is built upon a set of fixed
rules. We show via simulation studies that the Ci3+3 design exhibits similar
and comparable operating characteristics to more complex designs utilizing
model-based inferences. We believe that the Ci3+3 design may provide an
alternative choice to help simplify the design and conduct of combination
dose-finding trials in practice.
| stat |
FAVAE: Sequence Disentanglement using Information Bottleneck Principle | We propose the factorized action variational autoencoder (FAVAE), a
state-of-the-art generative model for learning disentangled and interpretable
representations from sequential data via the information bottleneck without
supervision. The purpose of disentangled representation learning is to obtain
interpretable and transferable representations from data. We focused on the
disentangled representation of sequential data since there is a wide range of
potential applications if disentanglement representation is extended to
sequential data such as video, speech, and stock market. Sequential data are
characterized by dynamic and static factors: dynamic factors are time
dependent, and static factors are independent of time. Previous models
disentangle static and dynamic factors by explicitly modeling the priors of
latent variables to distinguish between these factors. However, these models
cannot disentangle representations between dynamic factors, such as
disentangling "picking up" and "throwing" in robotic tasks. FAVAE can
disentangle multiple dynamic factors. Since it does not require modeling
priors, it can disentangle "between" dynamic factors. We conducted experiments
to show that FAVAE can extract disentangled dynamic factors.
| stat |
Estimation and Sensitivity Analysis for Causal Decomposition in Heath
Disparity Research | In the field of disparities research, there has been growing interest in
developing a counterfactual-based decomposition analysis to identify underlying
mediating mechanisms that help reduce disparities in populations. Despite rapid
development in the area, most prior studies have been limited to
regression-based methods, undermining the possibility of addressing complex
models with multiple mediators and/or heterogeneous effects. We propose an
estimation method that effectively addresses complex models. Moreover, we
develop a novel sensitivity analysis for possible violations of identification
assumptions. The proposed method and sensitivity analysis are demonstrated with
data from the Midlife Development in the US study to investigate the degree to
which disparities in cardiovascular health at the intersection of race and
gender would be reduced if the distributions of education and perceived
discrimination were the same across intersectional groups.
| stat |
Simple Sensitivity Analysis for Differential Measurement Error | Simple sensitivity analysis results are given for differential measurement
error of either the exposure or the outcome. In the case of differential
measurement error of the outcome it is shown that the true effect of the
exposure on the outcome on the risk ratio scale must be at least as large as
the observed association between the exposure and the mis-measured outcome
divided by the maximum strength of differential measurement error, assessed as
the risk ratio of the controlled direct effect of the exposure on mis-measured
outcome not through the true outcome. In the case of differential measurement
error of the exposure it is shown that the true effect on the risk ratio scale
of the exposure on the outcome must be at least as large as the observed
association between the mis-measured exposure measurement and the outcome
divided by the maximum strength of differential measurement error, assessed as
the risk ratio of the effect of the outcome on mis-measured exposure
measurement conditional on the true exposure. The results can also be
immediately used to indicate the minimum strength of differential measurement
error that would be needed to explain away an observed association between an
exposure measurement and an outcome measurement.
| stat |
Modeling Heterogeneity and Missing Data of Multiple Longitudinal
Outcomes in Electronic Health Records | In electronic health records (EHRs), latent subgroups of patients may exhibit
distinctive patterning in their longitudinal health trajectories. For such
data, growth mixture models (GMMs) enable classifying patients into different
latent classes based on individual trajectories and hypothesized risk factors.
However, the application of GMMs is hindered by the special missing data
problem in EHRs, which manifests two patient-led missing data processes: the
visit process and the response process for an EHR variable conditional on a
patient visiting the clinic. If either process is associated with the process
generating the longitudinal outcomes, then valid inferences require accounting
for a nonignorable missing data mechanism. We propose a Bayesian shared
parameter model that links GMMs of multiple longitudinal health outcomes, the
visit process, and the response process of each outcome given a visit using a
discrete latent class variable. Our focus is on multiple longitudinal health
outcomes for which there can be a clinically prescribed visit schedule. We
demonstrate our model in EHR measurements on early childhood weight and height
z-scores. Using data simulations, we illustrate the statistical properties of
our method with respect to subgroup-specific or marginal inferences. We built
the R package EHRMiss for model fitting, selection, and checking.
| stat |
Variance Networks: When Expectation Does Not Meet Your Expectations | Ordinary stochastic neural networks mostly rely on the expected values of
their weights to make predictions, whereas the induced noise is mostly used to
capture the uncertainty, prevent overfitting and slightly boost the performance
through test-time averaging. In this paper, we introduce variance layers, a
different kind of stochastic layers. Each weight of a variance layer follows a
zero-mean distribution and is only parameterized by its variance. We show that
such layers can learn surprisingly well, can serve as an efficient exploration
tool in reinforcement learning tasks and provide a decent defense against
adversarial attacks. We also show that a number of conventional Bayesian neural
networks naturally converge to such zero-mean posteriors. We observe that in
these cases such zero-mean parameterization leads to a much better training
objective than conventional parameterizations where the mean is being learned.
| stat |
Ultra high dimensional generalized additive model: Unified Theory and
Methods | Generalized additive model is a powerful statistical learning and predictive
modeling tool that has been applied in a wide range of applications. The need
of high-dimensional additive modeling is eminent in the context of dealing with
high through-put data such as genetic data analysis. In this article, we
studied a two step selection and estimation method for ultra high dimensional
generalized additive models. The first step applies group lasso on the expanded
bases of the functions. With high probability this selects all nonzero
functions without having too much over selection. The second step uses adaptive
group lasso with any initial estimators, including the group lasso estimator,
that satisfies some regular conditions. The adaptive group lasso estimator is
shown to be selection consistent with improved convergence rates. Tuning
parameter selection is also discussed and shown to select the true model
consistently under GIC procedure. The theoretical properties are supported by
extensive numerical study.
| stat |
Estimating high-resolution Red Sea surface temperature hotspots, using a
low-rank semiparametric spatial model | In this work, we estimate extreme sea surface temperature (SST) hotspots,
i.e., high threshold exceedance regions, for the Red Sea, a vital region of
high biodiversity. We analyze high-resolution satellite-derived SST data
comprising daily measurements at 16703 grid cells across the Red Sea over the
period 1985-2015. We propose a semiparametric Bayesian spatial mixed-effects
linear model with a flexible mean structure to capture spatially-varying trend
and seasonality, while the residual spatial variability is modeled through a
Dirichlet process mixture (DPM) of low-rank spatial Student-$t$ processes
(LTPs). By specifying cluster-specific parameters for each LTP mixture
component, the bulk of the SST residuals influence tail inference and hotspot
estimation only moderately. Our proposed model has a nonstationary mean,
covariance and tail dependence, and posterior inference can be drawn
efficiently through Gibbs sampling. In our application, we show that the
proposed method outperforms some natural parametric and semiparametric
alternatives. Moreover, we show how hotspots can be identified and we estimate
extreme SST hotspots for the whole Red Sea, projected until the year 2100,
based on the Representative Concentration Pathways 4.5 and 8.5. The estimated
95\% credible region for joint high threshold exceedances include large areas
covering major endangered coral reefs in the southern Red Sea.
| stat |
Best estimator for bivariate Poisson regression | INTRODUCTION: Wald's, the likelihood ratio (LR) and Rao's score tests and
their corresponding confidence intervals (CIs), are the three most common
estimators of parameters of Generalized Linear Models. On finite samples, these
estimators are biased. The objective of this work is to analyze the coverage
errors of the CI estimators in small samples for the log-Poisson model (i.e.
estimation of incidence rate ratio) with innovative evaluation criteria, taking
in account the overestimation/underestimation unbalance of coverage errors and
the variable inclusion rate and follow-up in epidemiological studies.
METHODS: Exact calculations equivalent to Monte Carlo simulations with an
infinite number of simulations have been used. Underestimation errors (due to
the upper bound of the CI) and overestimation coverage errors (due to the lower
bound of the CI) have been split. The level of confidence has been analyzed
from $0.95$ to $1-10^{-6}$, allowing the interpretation of P-values below
$10^{-6}$ for hypothesis tests.
RESULTS: The LR bias was small (actual coverage errors less than 1.5 times
the nominal errors) when the expected number of events in both groups was above
1, even when unbalanced (e.g. 10 events in one group vs 1 in the other). For
95% CI, Wald's and the Score estimators showed high bias even when the number
of events was large ($\geq 20$ in both groups) when groups were unbalanced. For
small P-values ($<10^{-6}$), the LR kept acceptable bias while Wald's and the
score P-values had severely inflated errors ($\times 100$).
CONCLUSION: The LR test and LR CI should be used.
| stat |
A non-autonomous equation discovery method for time signal
classification | Certain neural network architectures, in the infinite-layer limit, lead to
systems of nonlinear differential equations. Motivated by this idea, we develop
a framework for analyzing time signals based on non-autonomous dynamical
equations. We view the time signal as a forcing function for a dynamical system
that governs a time-evolving hidden variable. As in equation discovery, the
dynamical system is represented using a dictionary of functions and the
coefficients are learned from data. This framework is applied to the time
signal classification problem. We show how gradients can be efficiently
computed using the adjoint method, and we apply methods from dynamical systems
to establish stability of the classifier. Through a variety of experiments, on
both synthetic and real datasets, we show that the proposed method uses orders
of magnitude fewer parameters than competing methods, while achieving
comparable accuracy. We created the synthetic datasets using dynamical systems
of increasing complexity; though the ground truth vector fields are often
polynomials, we find consistently that a Fourier dictionary yields the best
results. We also demonstrate how the proposed method yields graphical
interpretability in the form of phase portraits.
| stat |
Blindness of score-based methods to isolated components and mixing
proportions | A large family of score-based methods are developed recently to solve
unsupervised learning problems including density estimation, statistical
testing and variational inference. These methods are attractive because they
exploit the derivative of the log density, which is independent of the
normaliser, and are thus suitable for tasks involving unnormalised densities.
Despite the theoretical guarantees on the performance, here we illustrate a
common practical issue suffered by these methods when the unnormalised
distribution of interest has isolated components. In particular, we study the
behaviour of some popular score-based methods on tasks involving 1-D mixture of
Gaussian. These methods fail to identify appropriate mixing proportions when
the unnormalised distribution is multimodal. Finally, some directions for
finding a remedy are discussed in light of recent successes in specific tasks.
We hope to bring the attention of theoreticians and practitioners to this issue
when developing new algorithms and applications.
| stat |
Failing Loudly: An Empirical Study of Methods for Detecting Dataset
Shift | We might hope that when faced with unexpected inputs, well-designed software
systems would fire off warnings. Machine learning (ML) systems, however, which
depend strongly on properties of their inputs (e.g. the i.i.d. assumption),
tend to fail silently. This paper explores the problem of building ML systems
that fail loudly, investigating methods for detecting dataset shift,
identifying exemplars that most typify the shift, and quantifying shift
malignancy. We focus on several datasets and various perturbations to both
covariates and label distributions with varying magnitudes and fractions of
data affected. Interestingly, we show that across the dataset shifts that we
explore, a two-sample-testing-based approach, using pre-trained classifiers for
dimensionality reduction, performs best. Moreover, we demonstrate that
domain-discriminating approaches tend to be helpful for characterizing shifts
qualitatively and determining if they are harmful.
| stat |
Riemannian Continuous Normalizing Flows | Normalizing flows have shown great promise for modelling flexible probability
distributions in a computationally tractable way. However, whilst data is often
naturally described on Riemannian manifolds such as spheres, torii, and
hyperbolic spaces, most normalizing flows implicitly assume a flat geometry,
making them either misspecified or ill-suited in these situations. To overcome
this problem, we introduce Riemannian continuous normalizing flows, a model
which admits the parametrization of flexible probability measures on smooth
manifolds by defining flows as the solution to ordinary differential equations.
We show that this approach can lead to substantial improvements on both
synthetic and real-world data when compared to standard flows or previously
introduced projected flows.
| stat |
On the sign recovery by LASSO, thresholded LASSO and thresholded Basis
Pursuit Denoising | We consider the regression model, when the number of observations is smaller
than the number of explicative variables. It is well known that the popular
Least Absolute Shrinkage and Selection Operator (LASSO) can recover the sign of
regression coefficients only if a very stringent irrepresentable condition is
satisfied. We extend this result by providing a tight upper bound for the
probability of LASSO sign recovery. The bound depends on the tuning parameter
and is attained when non-null components of the vector of regression
coefficients tend to infinity. In this situation it can be used to select the
value of the tuning parameter so as to control the probability of at least one
false discovery. Next, we revisit properties of thresholded LASSO and
thresholded Basis Pursuit Denoising (BPDN) and provide new theoretical results
in the asymptotic setup under which the design matrix is fixed and the
magnitudes of nonzero regression coefficients tend to infinity. We formulate an
easy identifiability condition which turns out to be sufficient and necessary
for thresholded LASSO and thresholded BPDN to recover the sign of the
sufficiently large signal. Our simulation study illustrates the large
difference between the irrepresentability and the identifiability condition,
especially when the entries in each row of the design matrix are strongly
correlated. Finally, we illustrate how the knockoff methodology allows to
select an appropriate threshold and that thresholded BPDN and thresholded LASSO
can recover the sign of the vector of regression coefficients with a larger
probability than adaptive LASSO.
| stat |
A High-Dimensional Particle Filter Algorithm | Online data assimilation in time series models over a large spatial extent is
an important problem in both geosciences and robotics. Such models are
intrinsically high-dimensional, rendering traditional particle filter
algorithms ineffective. Though methods that begin to address this problem
exist, they either rely on additional assumptions or lead to error that is
spatially inhomogeneous. I present a novel particle-based algorithm for online
approximation of the filtering problem on such models, using the fact that each
locus affects only nearby loci at the next time step. The algorithm is based on
a Metropolis-Hastings-like MCMC for creating hybrid particles at each step. I
show simulation results that suggest the error of this algorithm is uniform in
both space and time, with a lower bias, though higher variance, as compared to
a previously-proposed algorithm.
| stat |
Synthesizing Property & Casualty Ratemaking Datasets using Generative
Adversarial Networks | Due to confidentiality issues, it can be difficult to access or share
interesting datasets for methodological development in actuarial science, or
other fields where personal data are important. We show how to design three
different types of generative adversarial networks (GANs) that can build a
synthetic insurance dataset from a confidential original dataset. The goal is
to obtain synthetic data that no longer contains sensitive information but
still has the same structure as the original dataset and retains the
multivariate relationships. In order to adequately model the specific
characteristics of insurance data, we use GAN architectures adapted for
multi-categorical data: a Wassertein GAN with gradient penalty (MC-WGAN-GP), a
conditional tabular GAN (CTGAN) and a Mixed Numerical and Categorical
Differentially Private GAN (MNCDP-GAN). For transparency, the approaches are
illustrated using a public dataset, the French motor third party liability
data. We compare the three different GANs on various aspects: ability to
reproduce the original data structure and predictive models, privacy, and ease
of use. We find that the MC-WGAN-GP synthesizes the best data, the CTGAN is the
easiest to use, and the MNCDP-GAN guarantees differential privacy.
| stat |
Optimal Nested Simulation Experiment Design via Likelihood Ratio Method | Nested simulation arises frequently in financial or input uncertainty
quantification problems, where the performance measure is defined as a function
of the simulation output mean conditional on the outer scenario. The standard
nested simulation samples $M$ outer scenarios and runs $N$ inner replications
at each. We propose a new experiment design framework for a problem whose inner
replication's inputs are generated from probability distribution functions
parameterized by the outer scenario. This structure lets us pool replications
from an outer scenario to estimate another scenario's conditional mean via the
likelihood ratio method. We formulate a bi-level optimization problem to decide
not only which of $M$ outer scenarios to simulate and how many times to
replicate at each, but also how to pool these replications such that the total
simulation effort is minimized while achieving the same estimation error as the
standard nested simulation. The resulting optimal design requires far less
simulation effort than $MN$. We provide asymptotic analyses on the convergence
rates of the performance measure estimators computed from the experiment
design. Empirical results show that our experiment design significantly reduces
the simulation cost compared to the standard nested simulation as well as a
state-of-the-art design that pools replications via regressions.
| stat |
Robust Principal Component Analysis: A Median of Means Approach | Principal Component Analysis (PCA) is a fundamental tool for data
visualization, denoising, and dimensionality reduction. It is widely popular in
Statistics, Machine Learning, Computer Vision, and related fields. However, PCA
is well known to fall prey to the presence of outliers and often fails to
detect the true underlying low-dimensional structure within the dataset. Recent
supervised learning methods, following the Median of Means (MoM) philosophy,
have shown great success in dealing with outlying observations without much
compromise to their large sample theoretical properties. In this paper, we
propose a PCA procedure based on the MoM principle. Called the Median of Means
Principal Component Analysis (MoMPCA), the proposed method is not only
computationally appealing but also achieves optimal convergence rates under
minimal assumptions. In particular, we explore the non-asymptotic error bounds
of the obtained solution via the aid of Vapnik-Chervonenkis theory and
Rademacher complexity, while granting absolutely no assumption on the outlying
observations. The efficacy of the proposal is also thoroughly showcased through
simulations and real data applications.
| stat |
Should we Reload Time Series Classification Performance Evaluation ? (a
position paper) | Since the introduction and the public availability of the \textsc{ucr} time
series benchmark data sets, numerous Time Series Classification (TSC) methods
has been designed, evaluated and compared to each others. We suggest a critical
view of TSC performance evaluation protocols put in place in recent TSC
literature. The main goal of this `position' paper is to stimulate discussion
and reflexion about performance evaluation in TSC literature.
| stat |
Enabling Personalized Decision Support with Patient-Generated Data and
Attributable Components | Decision-making related to health is complex. Machine learning (ML) and
patient generated data can identify patterns and insights at the individual
level, where human cognition falls short, but not all ML-generated information
is of equal utility for making health-related decisions. We develop and apply
attributable components analysis (ACA), a method inspired by optimal transport
theory, to type 2 diabetes self-monitoring data to identify patterns of
association between nutrition and blood glucose control. In comparison with
linear regression, we found that ACA offers a number of characteristics that
make it promising for use in decision support applications. For example, ACA
was able to identify non-linear relationships, was more robust to outliers, and
offered broader and more expressive uncertainty estimates. In addition, our
results highlight a tradeoff between model accuracy and interpretability, and
we discuss implications for ML-driven decision support systems.
| stat |
Towards a global dynamic wind atlas: A multi-country validation of wind
power simulation from MERRA-2 and ERA-5 reanalyses bias-corrected with the
Global Wind Atlas | Reanalysis data are widely used for simulating renewable energy and in
particular wind power generation. While MERRA-2 has been a de-facto standard in
many studies, the newer ERA5- reanalysis recently gained importance. Here, we
use these two datasets to simulate wind power generation and evaluate the
respective quality in terms of correlations and errors when validated against
historical wind power generation. However, due to their coarse spatial
resolution, reanalyses fail to adequately represent local climatic conditions.
We therefore additionally apply mean bias correction with two versions of the
Global Wind Atlas (GWA) and assess the respective quality of resulting
simulations. Potential users of the dataset can also benefit from our analysis
of the impact of spatial and temporal aggregation on simulation quality
indicators. While similar studies have been conducted, they mainly cover
limited areas in Europe. In contrast, we look into regions, which globally
differ significantly in terms of the prevailing climate: the US, Brazil,
South-Africa, and New Zealand. Our principal findings are that (i) ERA5
outperforms MERRA-2, (ii) no major improvements can be expected by using
bias-correction with GWA2, while GWA3 even reduces simulation quality, and
(iii) temporal aggregation increases correlations and reduces errors, while
spatial aggregation does so only consistently when comparing very low and very
high aggregation levels.
| stat |
Global consensus Monte Carlo | To conduct Bayesian inference with large data sets, it is often convenient or
necessary to distribute the data across multiple machines. We consider a
likelihood function expressed as a product of terms, each associated with a
subset of the data. Inspired by global variable consensus optimisation, we
introduce an instrumental hierarchical model associating auxiliary statistical
parameters with each term, which are conditionally independent given the
top-level parameters. One of these top-level parameters controls the
unconditional strength of association between the auxiliary parameters. This
model leads to a distributed MCMC algorithm on an extended state space yielding
approximations of posterior expectations. A trade-off between computational
tractability and fidelity to the original model can be controlled by changing
the association strength in the instrumental model. We further propose the use
of a SMC sampler with a sequence of association strengths, allowing both the
automatic determination of appropriate strengths and for a bias correction
technique to be applied. In contrast to similar distributed Monte Carlo
algorithms, this approach requires few distributional assumptions. The
performance of the algorithms is illustrated with a number of simulated
examples.
| stat |
Likelihood-free approximate Gibbs sampling | Likelihood-free methods such as approximate Bayesian computation (ABC) have
extended the reach of statistical inference to problems with computationally
intractable likelihoods. Such approaches perform well for small-to-moderate
dimensional problems, but suffer a curse of dimensionality in the number of
model parameters. We introduce a likelihood-free approximate Gibbs sampler that
naturally circumvents the dimensionality issue by focusing on lower-dimensional
conditional distributions. These distributions are estimated by flexible
regression models either before the sampler is run, or adaptively during
sampler implementation. As a result, and in comparison to Metropolis-Hastings
based approaches, we are able to fit substantially more challenging statistical
models than would otherwise be possible. We demonstrate the sampler's
performance via two simulated examples, and a real analysis of Airbnb rental
prices using a intractable high-dimensional multivariate non-linear state space
model containing 13,140 parameters, which presents a real challenge to standard
ABC techniques.
| stat |
An Algorithmic Inference Approach to Learn Copulas | We introduce a new method for estimating the parameter of the bivariate
Clayton copulas within the framework of Algorithmic Inference. The method
consists of a variant of the standard boot-strapping procedure for inferring
random parameters, which we expressly devise to bypass the two pitfalls of this
specific instance: the non independence of the Kendall statistics, customarily
at the basis of this inference task, and the absence of a sufficient statistic
w.r.t. \alpha. The variant is rooted on a numerical procedure in order to find
the \alpha estimate at a fixed point of an iterative routine. Although paired
with the customary complexity of the program which computes them, numerical
results show an outperforming accuracy of the estimates.
| stat |
Improving the Reconstruction of Disentangled Representation Learners via
Multi-Stage Modelling | Current autoencoder-based disentangled representation learning methods
achieve disentanglement by penalizing the (aggregate) posterior to encourage
statistical independence of the latent factors. This approach introduces a
trade-off between disentangled representation learning and reconstruction
quality since the model does not have enough capacity to learn correlated
latent variables that capture detail information present in most image data. To
overcome this trade-off, we present a novel multi-stage modelling approach
where the disentangled factors are first learned using a preexisting
disentangled representation learning method (such as $\beta$-TCVAE); then, the
low-quality reconstruction is improved with another deep generative model that
is trained to model the missing correlated latent variables, adding detail
information while maintaining conditioning on the previously learned
disentangled factors. Taken together, our multi-stage modelling approach
results in a single, coherent probabilistic model that is theoretically
justified by the principal of D-separation and can be realized with a variety
of model classes including likelihood-based models such as variational
autoencoders, implicit models such as generative adversarial networks, and
tractable models like normalizing flows or mixtures of Gaussians. We
demonstrate that our multi-stage model has much higher reconstruction quality
than current state-of-the-art methods with equivalent disentanglement
performance across multiple standard benchmarks.
| stat |
Towards the interpretation of time-varying regularization parameters in
streaming penalized regression models | High-dimensional, streaming datasets are ubiquitous in modern applications.
Examples range from finance and e-commerce to the study of biomedical and
neuroimaging data. As a result, many novel algorithms have been proposed to
address challenges posed by such datasets. In this work, we focus on the use of
$\ell_1$ regularized linear models in the context of (possibly non-stationary)
streaming data Recently, it has been noted that the choice of the
regularization parameter is fundamental in such models and several methods have
been proposed which iteratively tune such a parameter in a~time-varying manner;
thereby allowing the underlying sparsity of estimated models to vary. Moreover,
in many applications, inference on the regularization parameter may itself be
of interest, as such a parameter is related to the underlying \textit{sparsity}
of the model. However, in this work, we highlight and provide extensive
empirical evidence regarding how various (often unrelated) statistical
properties in the data can lead to changes in the regularization parameter. In
particular, through various synthetic experiments, we demonstrate that changes
in the regularization parameter may be driven by changes in the true underlying
sparsity, signal-to-noise ratio or even model misspecification. The purpose of
this letter is, therefore, to highlight and catalog various statistical
properties which induce changes in the associated regularization parameter. We
conclude by presenting two applications: one relating to financial data and
another to neuroimaging data, where the aforementioned discussion is relevant.
| stat |
Revisiting the balance heuristic for estimating normalising constants | Multiple importance sampling estimators are widely used for computing
intractable constants due to its reliability and robustness. The celebrated
balance heuristic estimator belongs to this class of methods and has proved
very successful in computer graphics. The basic ingredients for computing the
estimator are: a set of proposal distributions, indexed by some discrete label,
and a predetermined number of draws from each of these proposals. However, if
the number of available proposals is much larger than the number of permitted
importance points, one needs to select, possibly at random, which of these
distributions will be used. The focus of this work lies within the previous
context, exploring some improvements and variations of the balance heuristic
via a novel extended-space representation of the estimator, leading to
straightforward annealing schemes for variance reduction purposes. In addition,
we also look at the intractable scenario where the proposal density is only
available as a joint function with the discrete label, as may be encountered in
problems where an ordering is imposed. For this case, we look at combinations
of correlated unbiased estimators which also fit into the extended-space
representation and, in turn, will provide other interesting solutions.
| stat |
Infinite Arms Bandit: Optimality via Confidence Bounds | Berry et al. (1997) initiated the development of the infinite arms bandit
problem. They derived a regret lower bound of all allocation strategies for
Bernoulli rewards with uniform priors, and proposed strategies based on success
runs. Bonald and Prouti\`{e}re (2013) proposed a two-target algorithm that
achieves the regret lower bound, and extended optimality to Bernoulli rewards
with general priors. We present here a confidence bound target (CBT) algorithm
that achieves optimality for rewards that are bounded above. For each arm we
construct a confidence bound and compare it against each other and a target
value to determine if the arm should be sampled further. The target value
depends on the assumed priors of the arm means. In the absence of information
on the prior, the target value is determined empirically. Numerical studies
here show that CBT is versatile and outperforms its competitors.
| stat |
Distribution Calibration for Regression | We are concerned with obtaining well-calibrated output distributions from
regression models. Such distributions allow us to quantify the uncertainty that
the model has regarding the predicted target value. We introduce the novel
concept of distribution calibration, and demonstrate its advantages over the
existing definition of quantile calibration. We further propose a post-hoc
approach to improving the predictions from previously trained regression
models, using multi-output Gaussian Processes with a novel Beta link function.
The proposed method is experimentally verified on a set of common regression
models and shows improvements for both distribution-level and quantile-level
calibration.
| stat |
Efficient Bayesian PARCOR Approaches for Dynamic Modeling of
Multivariate Time Series | A Bayesian lattice filtering and smoothing approach is proposed for fast and
accurate modeling and inference in multivariate non-stationary time series.
This approach offers computational feasibility and interpretable time-frequency
analysis in the multivariate context. The proposed framework allows us to
obtain posterior estimates of the time-varying spectral densities of individual
time series components, as well as posterior measurements of the time-frequency
relationships across multiple components, such as time-varying coherence and
partial coherence.
The proposed formulation considers multivariate dynamic linear models (MDLMs)
on the forward and backward time-varying partial autocorrelation coefficients
(TV-VPARCOR). Computationally expensive schemes for posterior inference on the
multivariate dynamic PARCOR model are avoided using approximations in the MDLM
context. Approximate inference on the corresponding time-varying vector
autoregressive (TV-VAR) coefficients is obtained via Whittle's algorithm. A key
aspect of the proposed TV-VPARCOR representations is that they are of lower
dimension, and therefore more efficient, than TV-VAR representations. The
performance of the TV-VPARCOR models is illustrated in simulation studies and
in the analysis of multivariate non-stationary temporal data arising in
neuroscience and environmental applications. Model performance is evaluated
using goodness-of-fit measurements in the time-frequency domain and also by
assessing the quality of short-term forecasting.
| stat |
The Effect of Medicaid Expansion on Non-Elderly Adult Uninsurance Rates
Among States that did not Expand Medicaid | We estimate the effect of Medicaid expansion on the adult uninsurance rate in
states that did not expand Medicaid in 2014. Using data from the American
Community Survey (ACS), we estimate this effect - the treatment effect on the
controls (ETC) - by re-weighting expansion regions to approximately balance the
covariates from non-expansion regions using an extension of the stable
balancing weights objective function (Zubizaretta (2015)). We contribute to the
balancing weights literature by accounting for hierarchical data structure and
covariate measurement error when calculating our weights, and to the synthetic
controls literature (see, e.g. Abadie et al 2010) by outlining a set of
assumptions that identifies the ETC using time-series cross-sectional data. We
estimate that Medicaid expansion would have changed the uninsurance rate by
-2.33 (-3.49, -1.16) percentage points. These results are smaller in absolute
magnitude than existing estimates of the treatment effect on the treated (ETT),
though may not be directly comparable due to the study design, target
population, and level of analysis. Regardless, we caution against making
inferences about the ETC using estimates of the ETT, and emphasize the need to
directly estimate the appropriate counterfactual when they are the quantity of
interest.
| stat |
Bayesian Deep Ensembles via the Neural Tangent Kernel | We explore the link between deep ensembles and Gaussian processes (GPs)
through the lens of the Neural Tangent Kernel (NTK): a recent development in
understanding the training dynamics of wide neural networks (NNs). Previous
work has shown that even in the infinite width limit, when NNs become GPs,
there is no GP posterior interpretation to a deep ensemble trained with squared
error loss. We introduce a simple modification to standard deep ensembles
training, through addition of a computationally-tractable, randomised and
untrainable function to each ensemble member, that enables a posterior
interpretation in the infinite width limit. When ensembled together, our
trained NNs give an approximation to a posterior predictive distribution, and
we prove that our Bayesian deep ensembles make more conservative predictions
than standard deep ensembles in the infinite width limit. Finally, using finite
width NNs we demonstrate that our Bayesian deep ensembles faithfully emulate
the analytic posterior predictive when available, and can outperform standard
deep ensembles in various out-of-distribution settings, for both regression and
classification tasks.
| stat |
Automatic model training under restrictive time constraints | We develop a hyperparameter optimisation algorithm, Automated Budget
Constrained Training (AutoBCT), which balances the quality of a model with the
computational cost required to tune it. The relationship between
hyperparameters, model quality and computational cost must be learnt and this
learning is incorporated directly into the optimisation problem. At each
training epoch, the algorithm decides whether to terminate or continue
training, and, in the latter case, what values of hyperparameters to use. This
decision weighs optimally potential improvements in the quality with the
additional training time and the uncertainty about the learnt quantities. The
performance of our algorithm is verified on a number of machine learning
problems encompassing random forests and neural networks. Our approach is
rooted in the theory of Markov decision processes with partial information and
we develop a numerical method to compute the value function and an optimal
strategy.
| stat |
Improving living biomass C-stock loss estimates by combining optical
satellite, airborne laser scanning, and NFI data | Policy measures and management decisions aiming at enhancing the role of
forests in mitigating climate-change require reliable estimates of C-stock
dynamics in greenhouse gas inventories (GHGIs). Aim of this study was to
assemble design-based estimators to provide estimates relevant for GHGIs using
national forest inventory (NFI) data. We improve basic expansion (BE) estimates
of living-biomass C-stock loss using field-data only, by leveraging with
remotely-sensed auxiliary data in model-assisted (MA) estimates. Our case
studies from Norway, Sweden, Denmark, and Latvia covered an area of >70 Mha.
Landsat-based Forest Cover Loss (FCL) and one-time wall-to-wall airborne laser
scanning (ALS) data served as auxiliary data. ALS provided information on the
C-stock before a potential disturbance indicated by FCL. The use of FCL in MA
estimators resulted in considerable efficiency gains which in most cases were
further increased by using ALS in addition. A doubling of efficiency was
possible for national estimates and even larger efficiencies were observed at
the sub-national level. Average annual estimates were considerably more precise
than pooled estimates using NFI data from all years at once. The combination of
remotely-sensed with NFI field data yields reliable estimates which is not
necessarily the case when using remotely-sensed data without reference
observations.
| stat |
Long-term Spatial Modeling for Characteristics of Extreme Heat Events | There is increasing evidence that global warming manifests itself in more
frequent warm days and that heat waves will become more frequent. Presently, a
formal definition of a heat wave is not agreed upon in the literature. To avoid
this debate, we consider extreme heat events, which, at a given location, are
well-defined as a run of consecutive days above an associated local threshold.
Characteristics of EHEs are of primary interest, such as incidence and
duration, as well as the magnitude of the average exceedance and maximum
exceedance above the threshold during the EHE.
Using approximately 60-year time series of daily maximum temperature data
collected at 18 locations in a given region, we propose a spatio-temporal model
to study the characteristics of EHEs over time. The model enables prediction of
the behavior of EHE characteristics at unobserved locations within the region.
Specifically, our approach employs a two-state space-time model for EHEs with
local thresholds where one state defines above threshold daily maximum
temperatures and the other below threshold temperatures. We show that our model
is able to recover the EHE characteristics of interest and outperforms a
corresponding autoregressive model that ignores thresholds based on
out-of-sample prediction.
| stat |
Differential analysis in Transcriptomic: The strength of randomly
picking 'reference' genes | Transcriptomic analysis are characterized by being not directly quantitative
and only providing relative measurements of expression levels up to an unknown
individual scaling factor. This difficulty is enhanced for differential
expression analysis. Several methods have been proposed to circumvent this lack
of knowledge by estimating the unknown individual scaling factors however, even
the most used one, are suffering from being built on hardly justifiable
biological hypotheses or from having weak statistical background. Only two
methods withstand this analysis: one based on largest connected graph component
hardly usable for large amount of expressions like in NGS, the second based on
$\log$-linear fits which unfortunately require a first step which uses one of
the methods described before.
We introduce a new procedure for differential analysis in the context of
transcriptomic data. It is the result of pooling together several differential
analyses each based on randomly picked genes used as reference genes. It
provides a differential analysis free from the estimation of the individual
scaling factors or any other knowledge. Theoretical properties are investigated
both in term of FWER and power. Moreover in the context of Poisson or negative
binomial modelization of the transcriptomic expressions, we derived a test with
non asymptotic control of its bounds. We complete our study by some empirical
simulations and apply our procedure to a real data set of hepatic miRNA
expressions from a mouse model of non-alcoholic steatohepatitis (NASH), the
CDAHFD model. This study on real data provides new hits with good biological
explanations.
| stat |
Multiscale Gaussian Process Level Set Estimation | In this paper, the problem of estimating the level set of a black-box
function from noisy and expensive evaluation queries is considered. A new
algorithm for this problem in the Bayesian framework with a Gaussian Process
(GP) prior is proposed. The proposed algorithm employs a hierarchical sequence
of partitions to explore different regions of the search space at varying
levels of detail depending upon their proximity to the level set boundary. It
is shown that this approach results in the algorithm having a low complexity
implementation whose computational cost is significantly smaller than the
existing algorithms for higher dimensional search space $\X$. Furthermore, high
probability bounds on a measure of discrepancy between the estimated level set
and the true level set for the the proposed algorithm are obtained, which are
shown to be strictly better than the existing guarantees for a large class of
GPs. In the process, a tighter characterization of the information gain of the
proposed algorithm is obtained which takes into account the structured nature
of the evaluation points. This approach improves upon the existing technique of
bounding the information gain with maximum information gain.
| stat |
ABCMETAapp: R Shiny Application for Simulation-based Estimation of Mean
and Standard Deviation for Meta-analysis via Approximate Bayesian Computation
(ABC) | Background and Objective: In meta-analysis based on continuous outcome,
estimated means and corresponding standard deviations from the selected studies
are key inputs to obtain a pooled estimate of the mean and its confidence
interval. We often encounter the situation that these quantities are not
directly reported in the literatures. Instead, other summary statistics are
reported such as median, minimum, maximum, quartiles, and study sample size.
Based on available summary statistics, we need to estimate estimates of mean
and standard deviation for meta-analysis. Methods: We developed a R Shiny code
based on approximate Bayesian computation (ABC), ABCMETA, to deal with this
situation. Results: In this article, we present an interactive and
user-friendly R Shiny application for implementing the proposed method (named
ABCMETAapp). In ABCMETAapp, users can choose an underlying outcome distribution
other than the normal distribution when the distribution of the outcome
variable is skewed or heavy tailed. We show how to run ABCMETAapp with
examples. Conclusions: ABCMETAapp provides a R Shiny implementation. This
method is more flexible than the existing analytical methods since estimation
can be based on five different distribution (Normal, Lognormal, Exponential,
Weibull, and Beta) for the outcome variable.
| stat |
A selective review on calibration information from similar studies based
on parametric likelihood or empirical likelihood | In multi-center clinical trials, due to various reasons, the individual-level
data are strictly restricted to be assessed publicly. Instead, the summarized
information is widely available from published results. With the advance of
computational technology, it has become very common in data analyses to run on
hundreds or thousands of machines simultaneous, with the data distributed
across those machines and no longer available in a single central location. How
to effectively assemble the summarized clinical data information or information
from each machine in parallel computation has become a challenging task for
statisticians and computer scientists. In this paper, we selectively review
some recently-developed statistical methods, including communication efficient
distributed statistical inference, and renewal estimation and incremental
inference, which can be regarded as the latest development of calibration
information methods in the era of big data. Even though those methods were
developed in different fields and in different statistical frameworks, in
principle, they are asymptotically equivalent to those well known methods
developed in meta analysis. Almost no or little information is lost compared
with the case when full data are available. As a general tool to integrate
information, we also review the generalized method of moments and estimating
equations approach by using empirical likelihood method.
| stat |
Estimation of subgraph density in noisy networks | While it is common practice in applied network analysis to report various
standard network summary statistics, these numbers are rarely accompanied by
uncertainty quantification. Yet any error inherent in the measurements
underlying the construction of the network, or in the network construction
procedure itself, necessarily must propagate to any summary statistics
reported. Here we study the problem of estimating the density of an arbitrary
subgraph, given a noisy version of some underlying network as data. Under a
simple model of network error, we show that consistent estimation of such
densities is impossible when the rates of error are unknown and only a single
network is observed. Accordingly, we develop method-of-moment estimators of
network subgraph densities and error rates for the case where a minimal number
of network replicates are available. These estimators are shown to be
asymptotically normal as the number of vertices increases to infinity. We also
provide confidence intervals for quantifying the uncertainty in these estimates
based on the asymptotic normality. To construct the confidence intervals, a new
and non-standard bootstrap method is proposed to compute asymptotic variances,
which is infeasible otherwise. We illustrate the proposed methods in the
context of gene coexpression networks.
| stat |
Forecasting model based on information-granulated GA-SVR and ARIMA for
producer price index | The accuracy of predicting the Producer Price Index (PPI) plays an
indispensable role in government economic work. However, it is difficult to
forecast the PPI. In our research, we first propose an unprecedented hybrid
model based on fuzzy information granulation that integrates the GA-SVR and
ARIMA (Autoregressive Integrated Moving Average Model) models. The
fuzzy-information-granulation-based GA-SVR-ARIMA hybrid model is intended to
deal with the problem of imprecision in PPI estimation. The proposed model
adopts the fuzzy information-granulation algorithm to
pre-classification-process monthly training samples of the PPI, and produced
three different sequences of fuzzy information granules, whose Support Vector
Regression (SVR) machine forecast models were separately established for their
Genetic Algorithm (GA) optimization parameters. Finally, the residual errors of
the GA-SVR model were rectified through ARIMA modeling, and the PPI estimate
was reached. Research shows that the PPI value predicted by this hybrid model
is more accurate than that predicted by other models, including ARIMA, GRNN,
and GA-SVR, following several comparative experiments. Research also indicates
the precision and validation of the PPI prediction of the hybrid model and
demonstrates that the model has consistent ability to leverage the forecasting
advantage of GA-SVR in non-linear space and of ARIMA in linear space.
| stat |
Estimation of Semi-Markov Multi-state Models: A Comparison of the
Sojourn Times and Transition Intensities Approaches | Semi-Markov models are widely used for survival analysis and reliability
analysis. In general, there are two competing parameterizations and each
entails its own interpretation and inference properties. On the one hand, a
semi-Markov process can be defined based on the distribution of sojourn times,
often via hazard rates, together with transition probabilities of an embedded
Markov chain. On the other hand, intensity transition functions may be used,
often referred to as the hazard rates of the semi-Markov process. We summarize
and contrast these two parameterizations both from a probabilistic and an
inference perspective, and we highlight relationships between the two
approaches. In general, the intensity transition based approach allows the
likelihood to be split into likelihoods of two-state models having fewer
parameters, allowing efficient computation and usage of many survival analysis
tools. {Nevertheless, in certain cases the sojourn time based approach is
natural and has been exploited extensively in applications.} In contrasting the
two approaches and contemporary relevant R packages used for inference, we use
two real datasets highlighting the probabilistic and inference properties of
each approach. This analysis is accompanied by an R vignette.
| stat |
NVAE: A Deep Hierarchical Variational Autoencoder | Normalizing flows, autoregressive models, variational autoencoders (VAEs),
and deep energy-based models are among competing likelihood-based frameworks
for deep generative learning. Among them, VAEs have the advantage of fast and
tractable sampling and easy-to-access encoding networks. However, they are
currently outperformed by other models such as normalizing flows and
autoregressive models. While the majority of the research in VAEs is focused on
the statistical challenges, we explore the orthogonal direction of carefully
designing neural architectures for hierarchical VAEs. We propose Nouveau VAE
(NVAE), a deep hierarchical VAE built for image generation using depth-wise
separable convolutions and batch normalization. NVAE is equipped with a
residual parameterization of Normal distributions and its training is
stabilized by spectral regularization. We show that NVAE achieves
state-of-the-art results among non-autoregressive likelihood-based models on
the MNIST, CIFAR-10, CelebA 64, and CelebA HQ datasets and it provides a strong
baseline on FFHQ. For example, on CIFAR-10, NVAE pushes the state-of-the-art
from 2.98 to 2.91 bits per dimension, and it produces high-quality images on
CelebA HQ. To the best of our knowledge, NVAE is the first successful VAE
applied to natural images as large as 256$\times$256 pixels. The source code is
available at https://github.com/NVlabs/NVAE .
| stat |
The Study of Urban Residential's Public Space Activeness using
Space-centric Approach | With the advancement of the Internet of Things (IoT) and communication
platform, large scale sensor deployment can be easily implemented in an urban
city to collect various information. To date, there are only a handful of
research studies about understanding the usage of urban public spaces.
Leveraging IoT, various sensors have been deployed in an urban residential area
to monitor and study public space utilization patterns. In this paper, we
propose a data processing system to generate space-centric insights about the
utilization of an urban residential region of multiple points of interest
(PoIs) that consists of 190,000m$^2$ real estate. We identify the activeness of
each PoI based on the spectral clustering, and then study their corresponding
static features, which are composed of transportation, commercial facilities,
population density, along with other characteristics. Through the heuristic
features inferring, the residential density and commercial facilities are the
most significant factors affecting public place utilization.
| stat |
Expectation propagation on the diluted Bayesian classifier | Efficient feature selection from high-dimensional datasets is a very
important challenge in many data-driven fields of science and engineering. We
introduce a statistical mechanics inspired strategy that addresses the problem
of sparse feature selection in the context of binary classification by
leveraging a computational scheme known as expectation propagation (EP). The
algorithm is used in order to train a continuous-weights perceptron learning a
classification rule from a set of (possibly partly mislabeled) examples
provided by a teacher perceptron with diluted continuous weights. We test the
method in the Bayes optimal setting under a variety of conditions and compare
it to other state-of-the-art algorithms based on message passing and on
expectation maximization approximate inference schemes. Overall, our
simulations show that EP is a robust and competitive algorithm in terms of
variable selection properties, estimation accuracy and computational
complexity, especially when the student perceptron is trained from correlated
patterns that prevent other iterative methods from converging. Furthermore, our
numerical tests demonstrate that the algorithm is capable of learning online
the unknown values of prior parameters, such as the dilution level of the
weights of the teacher perceptron and the fraction of mislabeled examples,
quite accurately. This is achieved by means of a simple maximum likelihood
strategy that consists in minimizing the free energy associated with the EP
algorithm.
| stat |
Reducing bias in difference-in-differences models using entropy
balancing | This paper illustrates the use of entropy balancing in
difference-in-differences analyses when pre-intervention outcome trends suggest
a possible violation of the parallel trends assumption. We describe a set of
assumptions under which weighting to balance intervention and comparison groups
on pre-intervention outcome trends leads to consistent
difference-in-differences estimates even when pre-intervention outcome trends
are not parallel. Simulated results verify that entropy balancing of
pre-intervention outcomes trends can remove bias when the parallel trends
assumption is not directly satisfied, and thus may enable researchers to use
difference-in-differences designs in a wider range of observational settings
than previously acknowledged.
| stat |
Reservoir Computing meets Recurrent Kernels and Structured Transforms | Reservoir Computing is a class of simple yet efficient Recurrent Neural
Networks where internal weights are fixed at random and only a linear output
layer is trained. In the large size limit, such random neural networks have a
deep connection with kernel methods. Our contributions are threefold: a) We
rigorously establish the recurrent kernel limit of Reservoir Computing and
prove its convergence. b) We test our models on chaotic time series prediction,
a classic but challenging benchmark in Reservoir Computing, and show how the
Recurrent Kernel is competitive and computationally efficient when the number
of data points remains moderate. c) When the number of samples is too large, we
leverage the success of structured Random Features for kernel approximation by
introducing Structured Reservoir Computing. The two proposed methods, Recurrent
Kernel and Structured Reservoir Computing, turn out to be much faster and more
memory-efficient than conventional Reservoir Computing.
| stat |
Where Bayes tweaks Gauss: Conditionally Gaussian priors for stable
multi-dipole estimation | We present a very simple yet powerful generalization of a previously
described model and algorithm for estimation of multiple dipoles from
magneto/electro-encephalographic data. Specifically, the generalization
consists in the introduction of a log-uniform hyperprior on the standard
deviation of a set of conditionally linear/Gaussian variables. We use numerical
simulations and an experimental dataset to show that the approximation to the
posterior distribution remains extremely stable under a wide range of values of
the hyperparameter, virtually removing the dependence on the hyperparameter.
| stat |
Multiple repairable systems under dependent competing risks with
nonparametric Frailty | The aim of this article is to analyze data from multiple repairable systems
under the presence of dependent competing risks. In order to model this
dependence structure, we adopted the well-known shared frailty model. This
model provides a suitable theoretical basis for generating dependence between
the components failure times in the dependent competing risks model. It is
known that the dependence effect in this scenario influences the estimates of
the model parameters. Hence, under the assumption that the cause-specific
intensities follow a PLP, we propose a frailty-induced dependence approach to
incorporate the dependence among the cause-specific recurrent processes.
Moreover, the misspecification of the frailty distribution may lead to errors
when estimating the parameters of interest. Because of this, we considered a
Bayesian nonparametric approach to model the frailty density in order to offer
more flexibility and to provide consistent estimates for the PLP model, as well
as insights about heterogeneity among the systems. Both simulation studies and
real case studies are provided to illustrate the proposed approaches and
demonstrate their validity.
| stat |
Multiscale Influenza Forecasting | Influenza forecasting in the United States (US) is complex and challenging
for reasons including substantial spatial and temporal variability, nested
geographic scales of forecast interest, and heterogeneous surveillance
participation. Here we present a flexible influenza forecasting model called
Dante, a multiscale flu forecasting model that learns rather than prescribes
spatial, temporal, and surveillance data structure. Forecasts at the Health and
Human Services (HHS) regional and national scales are generated as linear
combinations of state forecasts with weights proportional to US Census
population estimates, resulting in coherent forecasts across nested geographic
scales. We retrospectively compare Dante's short-term and seasonal forecasts at
the state, regional, and national scales for the 2012 through 2017 flu seasons
in the US to the Dynamic Bayesian Model (DBM), a leading flu forecasting model.
Dante outperformed DBM for nearly all spatial units, flu seasons, geographic
scales, and forecasting targets. The improved performance is due to Dante
making forecasts, especially short-term forecasts, more confidently and
accurately than DBM, suggesting Dante's improved forecast scores will also
translate to more useful forecasts for the public health sector. Dante
participated in the prospective 2018/19 FluSight challenge hosted by the
Centers for Disease Control and Prevention and placed 1st in both the national
and regional competition and the state competition. The methodology
underpinning Dante can be used in other disease forecasting contexts where
nested geographic scales of interest exist.
| stat |
On a computationally-scalable sparse formulation of the multidimensional
and non-stationary maximum entropy principle | Data-driven modelling and computational predictions based on maximum entropy
principle (MaxEnt-principle) aim at finding as-simple-as-possible - but not
simpler then necessary - models that allow to avoid the data overfitting
problem. We derive a multivariate non-parametric and non-stationary formulation
of the MaxEnt-principle and show that its solution can be approximated through
a numerical maximisation of the sparse constrained optimization problem with
regularization. Application of the resulting algorithm to popular financial
benchmarks reveals memoryless models allowing for simple and qualitative
descriptions of the major stock market indexes data. We compare the obtained
MaxEnt-models to the heteroschedastic models from the computational
econometrics (GARCH, GARCH-GJR, MS-GARCH, GARCH-PML4) in terms of the model
fit, complexity and prediction quality. We compare the resulting model
log-likelihoods, the values of the Bayesian Information Criterion, posterior
model probabilities, the quality of the data autocorrelation function fits as
well as the Value-at-Risk prediction quality. We show that all of the
considered seven major financial benchmark time series (DJI, SPX, FTSE, STOXX,
SMI, HSI and N225) are better described by conditionally memoryless
MaxEnt-models with nonstationary regime-switching than by the common
econometric models with finite memory. This analysis also reveals a sparse
network of statistically-significant temporal relations for the positive and
negative latent variance changes among different markets. The code is provided
for open access.
| stat |
An Optimistic Acceleration of AMSGrad for Nonconvex Optimization | We propose a new variant of AMSGrad, a popular adaptive gradient based
optimization algorithm widely used for training deep neural networks. Our
algorithm adds prior knowledge about the sequence of consecutive mini-batch
gradients and leverages its underlying structure making the gradients
sequentially predictable. By exploiting the predictability and ideas from
optimistic online learning, the proposed algorithm can accelerate the
convergence and increase sample efficiency. After establishing a tighter upper
bound under some convexity conditions on the regret, we offer a complimentary
view of our algorithm which generalizes the offline and stochastic version of
nonconvex optimization. In the nonconvex case, we establish a non-asymptotic
convergence bound independently of the initialization. We illustrate the
practical speedup on several deep learning models via numerical experiments.
| stat |
On the Convergence of Langevin Monte Carlo: The Interplay between Tail
Growth and Smoothness | We study sampling from a target distribution ${\nu_* = e^{-f}}$ using the
unadjusted Langevin Monte Carlo (LMC) algorithm. For any potential function $f$
whose tails behave like ${\|x\|^\alpha}$ for ${\alpha \in [1,2]}$, and has
$\beta$-H\"older continuous gradient, we prove that ${\widetilde{\mathcal{O}}
\Big(d^{\frac{1}{\beta}+\frac{1+\beta}{\beta}(\frac{2}{\alpha} -
\boldsymbol{1}_{\{\alpha \neq 1\}})} \epsilon^{-\frac{1}{\beta}}\Big)}$ steps
are sufficient to reach the $\epsilon $-neighborhood of a $d$-dimensional
target distribution $\nu_*$ in KL-divergence. This convergence rate, in terms
of $\epsilon$ dependency, is not directly influenced by the tail growth rate
$\alpha$ of the potential function as long as its growth is at least linear,
and it only relies on the order of smoothness $\beta$. One notable consequence
of this result is that for potentials with Lipschitz gradient, i.e. $\beta=1$,
our rate recovers the best known rate
${\widetilde{\mathcal{O}}(d\epsilon^{-1})}$ which was established for strongly
convex potentials in terms of $\epsilon$ dependency, but we show that the same
rate is achievable for a wider class of potentials that are degenerately convex
at infinity. The growth rate $\alpha$ starts to have an effect on the
established rate in high dimensions where $d$ is large; furthermore, it
recovers the best-known dimension dependency when the tail growth of the
potential is quadratic, i.e. ${\alpha = 2}$, in the current setup. Our
framework allows for finite perturbations, and any order of smoothness
${\beta\in(0,1]}$; consequently, our results are applicable to a wide class of
non-convex potentials that are weakly smooth and exhibit at least linear tail
growth.
| stat |
The Universal model and prior: multinomial GLMs | This paper generalises the exponential family GLM to allow arbitrary
distributions for the response variable. This is achieved by combining the
model-assisted regression approach from survey sampling with the GLM scoring
algorithm, weighted by random draws from the posterior Dirichlet distribution
of the support point probabilities of the multinomial distribution. The
generalisation provides fully Bayesian analyses from the posterior sampling,
without MCMC. Several examples are given, of published GLM data sets. The
approach can be extended widely: an example of a GLMM extension is given.
| stat |
Inductive Mutual Information Estimation: A Convex Maximum-Entropy Copula
Approach | We propose a novel estimator of the mutual information between two ordinal
vectors $x$ and $y$. Our approach is inductive (as opposed to deductive) in
that it depends on the data generating distribution solely through some
nonparametric properties revealing associations in the data, and does not
require having enough data to fully characterize the true joint distributions
$P_{x, y}$. Specifically, our approach consists of (i) noting that $I\left(y;
x\right) = I\left(u_y; u_x\right)$ where $u_y$ and $u_x$ are the copula-uniform
dual representations of $y$ and $x$ (i.e. their images under the probability
integral transform), and (ii) estimating the copula entropies
$h\left(u_y\right)$, $h\left(u_x\right)$ and $h\left(u_y, u_x\right)$ by
solving a maximum-entropy problem over the space of copula densities under a
constraint of the type $\alpha_m = E\left[\phi_m(u_y, u_x)\right]$. We prove
that, so long as the constraint is feasible, this problem admits a unique
solution, it is in the exponential family, and it can be learned by solving a
convex optimization problem. The resulting estimator, which we denote MIND, is
marginal-invariant, always non-negative, unbounded for any sample size $n$,
consistent, has MSE rate $O(1/n)$, and is more data-efficient than competing
approaches. Beyond mutual information estimation, we illustrate that our
approach may be used to mitigate mode collapse in GANs by maximizing the
entropy of the copula of fake samples, a model we refer to as Copula Entropy
Regularized GAN (CER-GAN).
| stat |
A generalization of regularized dual averaging and its dynamics | Excessive computational cost for learning large data and streaming data can
be alleviated by using stochastic algorithms, such as stochastic gradient
descent and its variants. Recent advances improve stochastic algorithms on
convergence speed, adaptivity and structural awareness. However, distributional
aspects of these new algorithms are poorly understood, especially for
structured parameters. To develop statistical inference in this case, we
propose a class of generalized regularized dual averaging (gRDA) algorithms
with constant step size, which improves RDA (Xiao, 2010; Flammarion and Bach,
2017). Weak convergence of gRDA trajectories are studied, and as a consequence,
for the first time in the literature, the asymptotic distributions for online
l1 penalized problems become available. These general results apply to both
convex and non-convex differentiable loss functions, and in particular, recover
the existing regret bound for convex losses (Nemirovski et al., 2009). As
important applications, statistical inferential theory on online sparse linear
regression and online sparse principal component analysis are developed, and
are supported by extensive numerical analysis. Interestingly, when gRDA is
properly tuned, support recovery and central limiting distribution (with mean
zero) hold simultaneously in the online setting, which is in contrast with the
biased central limiting distribution of batch Lasso (Knight and Fu, 2000).
Technical devices, including weak convergence of stochastic mirror descent, are
developed as by-products with independent interest. Preliminary empirical
analysis of modern image data shows that learning very sparse deep neural
networks by gRDA does not necessarily sacrifice testing accuracy.
| stat |
An Empirical Bayes Approach to Graphon Estimation | The graphon (W-graph), including the stochastic block model as a special
case, has been widely used in modeling and analyzing network data. This random
graph model is well-characterized by its graphon function, and estimation of
the graphon function has gained a lot of recent research interests. Most
existing works focus on community detection in the latent space of the model,
while adopting simple maximum likelihood or Bayesian estimates for the graphon
or connectivity parameters given the identified communities. In this work, we
propose a hierarchical Binomial model and develop a novel empirical Bayes
estimate of the connectivity matrix of a stochastic block model to approximate
the graphon function. Based on the likelihood of our hierarchical model, we
further introduce a model selection criterion for choosing the number of
communities. Numerical results on extensive simulations and two well-annotated
social networks demonstrate the superiority of our approach in terms of
estimation accuracy and model selection.
| stat |
Gaussian DAGs on network data | The traditional directed acyclic graph (DAG) model assumes data are generated
independently from the underlying joint distribution defined by the DAG. In
many applications, however, individuals are linked via a network and thus the
independence assumption does not hold. We propose a novel Gaussian DAG model
for network data, where the dependence among individual data points (row
covariance) is modeled by an undirected graph. Under this model, we develop a
maximum penalized likelihood method to estimate the DAG structure and the row
correlation matrix. The algorithm iterates between a decoupled lasso regression
step and a graphical lasso step. We show with extensive simulated and real
network data, that our algorithm improves the accuracy of DAG structure
learning by leveraging the information from the estimated row correlations.
Moreover, we demonstrate that the performance of existing DAG learning methods
can be substantially improved via de-correlation of network data with the
estimated row correlation matrix from our algorithm.
| stat |
GAN-based Priors for Quantifying Uncertainty | Bayesian inference is used extensively to quantify the uncertainty in an
inferred field given the measurement of a related field when the two are linked
by a mathematical model. Despite its many applications, Bayesian inference
faces challenges when inferring fields that have discrete representations of
large dimension, and/or have prior distributions that are difficult to
characterize mathematically. In this work we demonstrate how the approximate
distribution learned by a deep generative adversarial network (GAN) may be used
as a prior in a Bayesian update to address both these challenges. We
demonstrate the efficacy of this approach on two distinct, and remarkably
broad, classes of problems. The first class leads to supervised learning
algorithms for image classification with superior out of distribution detection
and accuracy, and for image inpainting with built-in variance estimation. The
second class leads to unsupervised learning algorithms for image denoising and
for solving physics-driven inverse problems.
| stat |
A simple algorithm for global sensitivity analysis with Shapley effects | Global sensitivity analysis aims at measuring the relative importance of
different variables or groups of variables for the variability of a quantity of
interest. Among several sensitivity indices, so-called Shapley effects have
recently gained popularity mainly because the Shapley effects for all the
individual variables are summed up to the overall variance, which gives a
better interpretability than the classical sensitivity indices called main
effects and total effects. In this paper, assuming that all the input variables
are independent, we introduce a quite simple Monte Carlo algorithm to estimate
the Shapley effects for all the individual variables simultaneously, which
drastically simplifies the existing algorithms proposed in the literature. We
present a short Matlab implementation of our algorithm and show some numerical
results. A possible extension to the case where the input variables are
dependent is also discussed.
| stat |
Warped Input Gaussian Processes for Time Series Forecasting | We introduce a Gaussian process-based model for handling of non-stationarity.
The warping is achieved non-parametrically, through imposing a prior on the
relative change of distance between subsequent observation inputs. The model
allows the use of general gradient optimization algorithms for training and
incurs only a small computational overhead on training and prediction. The
model finds its applications in forecasting in non-stationary time series with
either gradually varying volatility, presence of change points, or a
combination thereof. We evaluate the model on synthetic and real-world time
series data comparing against both baseline and known state-of-the-art
approaches and show that the model exhibits state-of-the-art forecasting
performance at a lower implementation and computation cost.
| stat |
State Heterogeneity Analysis of Financial Volatility Using
High-Frequency Financial Data | Recently, to account for low-frequency market dynamics, several volatility
models, employing high-frequency financial data, have been developed. However,
in financial markets, we often observe that financial volatility processes
depend on economic states, so they have a state heterogeneous structure. In
this paper, to study state heterogeneous market dynamics based on
high-frequency data, we introduce a novel volatility model based on a
continuous Ito diffusion process whose intraday instantaneous volatility
process evolves depending on the exogenous state variable, as well as its
integrated volatility. We call it the state heterogeneous GARCH-Ito (SG-Ito)
model. We suggest a quasi-likelihood estimation procedure with the realized
volatility proxy and establish its asymptotic behaviors. Moreover, to test the
low-frequency state heterogeneity, we develop a Wald test-type hypothesis
testing procedure. The results of empirical studies suggest the existence of
leverage, investor attention, market illiquidity, stock market comovement, and
post-holiday effect in S&P 500 index volatility.
| stat |
Convex Fairness Constrained Model Using Causal Effect Estimators | Recent years have seen much research on fairness in machine learning. Here,
mean difference (MD) or demographic parity is one of the most popular measures
of fairness. However, MD quantifies not only discrimination but also
explanatory bias which is the difference of outcomes justified by explanatory
features. In this paper, we devise novel models, called FairCEEs, which remove
discrimination while keeping explanatory bias. The models are based on
estimators of causal effect utilizing propensity score analysis. We prove that
FairCEEs with the squared loss theoretically outperform a naive MD constraint
model. We provide an efficient algorithm for solving FairCEEs in regression and
binary classification tasks. In our experiment on synthetic and real-world data
in these two tasks, FairCEEs outperformed an existing model that considers
explanatory bias in specific cases.
| stat |
Imbalanced classification: an objective-oriented review | A common issue for classification in scientific research and industry is the
existence of imbalanced classes. When sample sizes of different classes are
imbalanced in training data, naively implementing a classification method often
leads to unsatisfactory prediction results on test data. Multiple resampling
techniques have been proposed to address the class imbalance issues. Yet, there
is no general guidance on when to use each technique. In this article, we
provide an objective-oriented review of the common resampling techniques for
binary classification under imbalanced class sizes. The learning objectives we
consider include the classical paradigm that minimizes the overall
classification error, the cost-sensitive learning paradigm that minimizes a
cost-adjusted weighted type I and type II errors, and the Neyman-Pearson
paradigm that minimizes the type II error subject to a type I error constraint.
Under each paradigm, we investigate the combination of the resampling
techniques and a few state-of-the-art classification methods. For each pair of
resampling techniques and classification methods, we use simulation studies to
study the performance under different evaluation metrics. From these extensive
simulation experiments, we demonstrate under each classification paradigm, the
complex dynamics among resampling techniques, base classification methods,
evaluation metrics, and imbalance ratios. For practitioners, the take-away
message is that with imbalanced data, one usually should consider all the
combinations of resampling techniques and the base classification methods.
| stat |
Optimal Projections in the Distance-Based Statistical Methods | This paper introduces a new way to calculate distance-based statistics,
particularly when the data are multivariate. The main idea is to pre-calculate
the optimal projection directions given the variable dimension, and to project
multidimensional variables onto these pre-specified projection directions; by
subsequently utilizing the fast algorithm that is developed in Huo and
Sz\'ekely [2016] for the univariate variables, the computational complexity can
be improved from $O(m^2)$ to $O(n m \cdot \mbox{log}(m))$, where $n$ is the
number of projection directions and $m$ is the sample size. When $n \ll
m/\log(m)$, computational savings can be achieved. The key challenge is how to
find the optimal pre-specified projection directions. This can be obtained by
minimizing the worse-case difference between the true distance and the
approximated distance, which can be formulated as a nonconvex optimization
problem in a general setting. In this paper, we show that the exact solution of
the nonconvex optimization problem can be derived in two special cases: the
dimension of the data is equal to either $2$ or the number of projection
directions. In the generic settings, we propose an algorithm to find some
approximate solutions. Simulations confirm the advantage of our method, in
comparison with the pure Monte Carlo approach, in which the directions are
randomly selected rather than pre-calculated.
| stat |
Copula Flows for Synthetic Data Generation | The ability to generate high-fidelity synthetic data is crucial when
available (real) data is limited or where privacy and data protection standards
allow only for limited use of the given data, e.g., in medical and financial
data-sets. Current state-of-the-art methods for synthetic data generation are
based on generative models, such as Generative Adversarial Networks (GANs).
Even though GANs have achieved remarkable results in synthetic data generation,
they are often challenging to interpret.Furthermore, GAN-based methods can
suffer when used with mixed real and categorical variables.Moreover, loss
function (discriminator loss) design itself is problem specific, i.e., the
generative model may not be useful for tasks it was not explicitly trained for.
In this paper, we propose to use a probabilistic model as a synthetic data
generator. Learning the probabilistic model for the data is equivalent to
estimating the density of the data. Based on the copula theory, we divide the
density estimation task into two parts, i.e., estimating univariate marginals
and estimating the multivariate copula density over the univariate marginals.
We use normalising flows to learn both the copula density and univariate
marginals. We benchmark our method on both simulated and real data-sets in
terms of density estimation as well as the ability to generate high-fidelity
synthetic data
| stat |
Approximate Cross-validation: Guarantees for Model Assessment and
Selection | Cross-validation (CV) is a popular approach for assessing and selecting
predictive models. However, when the number of folds is large, CV suffers from
a need to repeatedly refit a learning procedure on a large number of training
datasets. Recent work in empirical risk minimization (ERM) approximates the
expensive refitting with a single Newton step warm-started from the full
training set optimizer. While this can greatly reduce runtime, several open
questions remain including whether these approximations lead to faithful model
selection and whether they are suitable for non-smooth objectives. We address
these questions with three main contributions: (i) we provide uniform
non-asymptotic, deterministic model assessment guarantees for approximate CV;
(ii) we show that (roughly) the same conditions also guarantee model selection
performance comparable to CV; (iii) we provide a proximal Newton extension of
the approximate CV framework for non-smooth prediction problems and develop
improved assessment guarantees for problems such as l1-regularized ERM.
| stat |
An n-dimensional Rosenbrock Distribution for MCMC Testing | The Rosenbrock function is an ubiquitous benchmark problem for numerical
optimisation, and variants have been proposed to test the performance of Markov
Chain Monte Carlo algorithms. In this work we discuss the two-dimensional
Rosenbrock density, its current $n$-dimensional extensions, and their
advantages and limitations. We then propose a new extension to arbitrary
dimensions called the Hybrid Rosenbrock distribution, which is composed of
conditional normal kernels arranged in such a way that preserves the key
features of the original kernel. Moreover, due to its structure, the Hybrid
Rosenbrock distribution is analytically tractable and possesses several
desirable properties, which make it an excellent test model for computational
algorithms.
| stat |
Learn-By-Calibrating: Using Calibration as a Training Objective | Calibration error is commonly adopted for evaluating the quality of
uncertainty estimators in deep neural networks. In this paper, we argue that
such a metric is highly beneficial for training predictive models, even when we
do not explicitly measure the uncertainties. This is conceptually similar to
heteroscedastic neural networks that produce variance estimates for each
prediction, with the key difference that we do not place a Gaussian prior on
the predictions. We propose a novel algorithm that performs simultaneous
interval estimation for different calibration levels and effectively leverages
the intervals to refine the mean estimates. Our results show that, our approach
is consistently superior to existing regularization strategies in deep
regression models. Finally, we propose to augment partial dependence plots, a
model-agnostic interpretability tool, with expected prediction intervals to
reveal interesting dependencies between data and the target.
| stat |
KL-UCB-switch: optimal regret bounds for stochastic bandits from both a
distribution-dependent and a distribution-free viewpoints | In the context of K-armed stochastic bandits with distribution only assumed
to be supported by [0,1], we introduce the first algorithm, called
KL-UCB-switch, that enjoys simultaneously a distribution-free regret bound of
optimal order $\sqrt{KT}$ and a distribution-dependent regret bound of optimal
order as well, that is, matching the $\kappa\ln T$ lower bound by Lai & Robbins
(1985) and Burnetas & Katehakis (1996). This self-contained contribution
simultaneously presents state-of-the-art techniques for regret minimization in
bandit models, and an elementary construction of non-asymptotic confidence
bounds based on the empirical likelihood method for bounded distributions.
| stat |
When does gradient descent with logistic loss interpolate using deep
networks with smoothed ReLU activations? | We establish conditions under which gradient descent applied to fixed-width
deep networks drives the logistic loss to zero, and prove bounds on the rate of
convergence. Our analysis applies for smoothed approximations to the ReLU, such
as Swish and the Huberized ReLU, proposed in previous applied work. We provide
two sufficient conditions for convergence. The first is simply a bound on the
loss at initialization. The second is a data separation condition used in prior
analyses.
| stat |
BOCK : Bayesian Optimization with Cylindrical Kernels | A major challenge in Bayesian Optimization is the boundary issue (Swersky,
2017) where an algorithm spends too many evaluations near the boundary of its
search space. In this paper, we propose BOCK, Bayesian Optimization with
Cylindrical Kernels, whose basic idea is to transform the ball geometry of the
search space using a cylindrical transformation. Because of the transformed
geometry, the Gaussian Process-based surrogate model spends less budget
searching near the boundary, while concentrating its efforts relatively more
near the center of the search region, where we expect the solution to be
located. We evaluate BOCK extensively, showing that it is not only more
accurate and efficient, but it also scales successfully to problems with a
dimensionality as high as 500. We show that the better accuracy and scalability
of BOCK even allows optimizing modestly sized neural network layers, as well as
neural network hyperparameters.
| stat |
Self-Regularity of Non-Negative Output Weights for Overparameterized
Two-Layer Neural Networks | We consider the problem of finding a two-layer neural network with sigmoid,
rectified linear unit (ReLU), or binary step activation functions that "fits" a
training data set as accurately as possible as quantified by the training
error; and study the following question: \emph{does a low training error
guarantee that the norm of the output layer (outer norm) itself is small?} We
answer affirmatively this question for the case of non-negative output weights.
Using a simple covering number argument, we establish that under quite mild
distributional assumptions on the input/label pairs; any such network achieving
a small training error on polynomially many data necessarily has a
well-controlled outer norm. Notably, our results (a) have a polynomial (in $d$)
sample complexity, (b) are independent of the number of hidden units (which can
potentially be very high), (c) are oblivious to the training algorithm; and (d)
require quite mild assumptions on the data (in particular the input vector
$X\in\mathbb{R}^d$ need not have independent coordinates). We then leverage our
bounds to establish generalization guarantees for such networks through
\emph{fat-shattering dimension}, a scale-sensitive measure of the complexity
class that the network architectures we investigate belong to. Notably, our
generalization bounds also have good sample complexity (polynomials in $d$ with
a low degree), and are in fact near-linear for some important cases of
interest.
| stat |
Design-unbiased statistical learning in survey sampling | Design-consistent model-assisted estimation has become the standard practice
in survey sampling. However, a general theory is lacking so far, which allows
one to incorporate modern machine-learning techniques that can lead to
potentially much more powerful assisting models. We propose a subsampling
Rao-Blackwell method, and develop a statistical learning theory for exactly
design-unbiased estimation with the help of linear or non-linear prediction
models. Our approach makes use of classic ideas from Statistical Science as
well as the rapidly growing field of Machine Learning. Provided rich auxiliary
information, it can yield considerable efficiency gains over standard linear
model-assisted methods, while ensuring valid estimation for the given target
population, which is robust against potential mis-specifications of the
assisting model at the individual level.
| stat |
Non-intrusive and semi-intrusive uncertainty quantification of a
multiscale in-stent restenosis model | Uncertainty estimations are presented of the response of a multiscale
in-stent restenosis model, as obtained by both non-intrusive and semi-intrusive
uncertainty quantification. The in-stent restenosis model is a fully coupled
multiscale simulation of post-stenting tissue growth, in which the most costly
submodel is the blood flow simulation. Surrogate modeling for non-intrusive
uncertainty quantification takes the whole model as a black-box and maps
directly from the three uncertain inputs to the quantity of interest, the
neointimal area. The corresponding uncertain estimates matched the results from
quasi-Monte Carlo simulations well. In the semi-intrusive uncertainty
quantification, the most expensive submodel is replaced with a surrogate model.
We developed a surrogate model for the blood flow simulation by using a
convolutional neural network. The semi-intrusive method with the new surrogate
model offered efficient estimates of uncertainty and sensitivity while keeping
relatively high accuracy. It outperformed the result obtained with earlier
surrogate models. It also achieved the estimates comparable to the
non-intrusive method with similar efficiency. Presented results on uncertainty
propagation with non-intrusive and semi-intrusive metamodeling methods allow us
to draw some conclusions on the advantages and limitations of these methods.
| stat |
Spectral embedding of weighted graphs | This paper concerns the statistical analysis of a weighted graph through
spectral embedding. Under a latent position model in which the expected
adjacency matrix has low rank, we prove uniform consistency and a central limit
theorem for the embedded nodes, treated as latent position estimates. In the
special case of a weighted stochastic block model, this result implies that the
embedding follows a Gaussian mixture model with each component representing a
community. We exploit this to formally evaluate different weight
representations of the graph using Chernoff information. For example, in a
network anomaly detection problem where we observe a p-value on each edge, we
recommend against directly embedding the matrix of p-values, and instead using
threshold or log p-values, depending on network sparsity and signal strength.
| stat |
Efficient estimation of the ANOVA mean dimension, with an application to
neural net classification | The mean dimension of a black box function of $d$ variables is a convenient
way to summarize the extent to which it is dominated by high or low order
interactions. It is expressed in terms of $2^d-1$ variance components but it
can be written as the sum of $d$ Sobol' indices that can be estimated by leave
one out methods. We compare the variance of these leave one out methods: a
Gibbs sampler called winding stairs, a radial sampler that changes each
variable one at a time from a baseline, and a naive sampler that never reuses
function evaluations and so costs about double the other methods. For an
additive function the radial and winding stairs are most efficient. For a
multiplicative function the naive method can easily be most efficient if the
factors have high kurtosis. As an illustration we consider the mean dimension
of a neural network classifier of digits from the MNIST data set. The
classifier is a function of $784$ pixels. For that problem, winding stairs is
the best algorithm. We find that inputs to the final softmax layer have mean
dimensions ranging from $1.35$ to $2.0$.
| stat |
Variational Bayesian Decision-making for Continuous Utilities | Bayesian decision theory outlines a rigorous framework for making optimal
decisions based on maximizing expected utility over a model posterior. However,
practitioners often do not have access to the full posterior and resort to
approximate inference strategies. In such cases, taking the eventual
decision-making task into account while performing the inference allows for
calibrating the posterior approximation to maximize the utility. We present an
automatic pipeline that co-opts continuous utilities into variational inference
algorithms to account for decision-making. We provide practical strategies for
approximating and maximizing the gain, and empirically demonstrate consistent
improvement when calibrating approximations for specific utilities.
| stat |
A Bayesian semi-parametric hybrid model for spatial extremes with
unknown dependence structure | The max-stable process is an asymptotically justified model for spatial
extremes. In particular, we focus on the hierarchical extreme-value process
(HEVP), which is a particular max-stable process that is conducive to Bayesian
computing. The HEVP and all max-stable process models are parametric and impose
strong assumptions including that all marginal distributions belong to the
generalized extreme value family and that nearby sites are asymptotically
dependent. We generalize the HEVP by relaxing these assumptions to provide a
wider class of marginal distributions via a Dirichlet process prior for the
spatial random effects distribution. In addition, we present a hybrid
max-mixture model that combines the strengths of the parametric and
semi-parametric models. We show that this versatile max-mixture model
accommodates both asymptotic independence and dependence and can be fit using
standard Markov chain Monte Carlo algorithms. The utility of our model is
evaluated in Monte Carlo simulation studies and application to Netherlands wind
gust data.
| stat |
Truncated Variational Expectation Maximization | We derive a novel variational expectation maximization approach based on
truncated posterior distributions. Truncated distributions are proportional to
exact posteriors within subsets of a discrete state space and equal zero
otherwise. The treatment of the distributions' subsets as variational
parameters distinguishes the approach from previous variational approaches. The
specific structure of truncated distributions allows for deriving novel and
mathematically grounded results, which in turn can be used to formulate novel
efficient algorithms to optimize the parameters of probabilistic generative
models. Most centrally, we find the variational lower bounds that correspond to
truncated distributions to be given by very concise and efficiently computable
expressions, while update equations for model parameters remain in their
standard form. Based on these findings, we show how efficient and easily
applicable meta-algorithms can be formulated that guarantee a monotonic
increase of the variational bound. Example applications of the here derived
framework provide novel theoretical results and learning procedures for latent
variable models as well as mixture models. Furthermore, we show that truncated
variation EM naturally interpolates between standard EM with full posteriors
and EM based on the maximum a-posteriori state (MAP). The approach can,
therefore, be regarded as a generalization of the popular `hard EM' approach
towards a similarly efficient method which can capture more of the true
posterior structure.
| stat |
Resource-Efficient Neural Networks for Embedded Systems | While machine learning is traditionally a resource intensive task, embedded
systems, autonomous navigation, and the vision of the Internet of Things fuel
the interest in resource-efficient approaches. These approaches aim for a
carefully chosen trade-off between performance and resource consumption in
terms of computation and energy. The development of such approaches is among
the major challenges in current machine learning research and key to ensure a
smooth transition of machine learning technology from a scientific environment
with virtually unlimited computing resources into every day's applications. In
this article, we provide an overview of the current state of the art of machine
learning techniques facilitating these real-world requirements. In particular,
we focus on deep neural networks (DNNs), the predominant machine learning
models of the past decade. We give a comprehensive overview of the vast
literature that can be mainly split into three non-mutually exclusive
categories: (i) quantized neural networks, (ii) network pruning, and (iii)
structural efficiency. These techniques can be applied during training or as
post-processing, and they are widely used to reduce the computational demands
in terms of memory footprint, inference speed, and energy efficiency. We
substantiate our discussion with experiments on well-known benchmark data sets
to showcase the difficulty of finding good trade-offs between
resource-efficiency and predictive performance.
| stat |
Adversarial training for predictive tasks: theoretical analysis and
limitations in the deterministic case | To train a deep neural network to mimic the outcomes of processing sequences,
a version of Conditional Generalized Adversarial Network (CGAN) can be used. It
has been observed by others that CGAN can help to improve the results even for
deterministic sequences, where only one output is associated with the
processing of a given input. Surprisingly, our CGAN-based tests on
deterministic geophysical processing sequences did not produce a real
improvement compared to the use of an $L_p$ loss; we here propose a first
theoretical explanation why. Our analysis goes from the non-deterministic case
to the deterministic one. It led us to develop an adversarial way to train a
content loss that gave better results on our data.
| stat |
Minibatch optimal transport distances; analysis and applications | Optimal transport distances have become a classic tool to compare probability
distributions and have found many applications in machine learning. Yet,
despite recent algorithmic developments, their complexity prevents their direct
use on large scale datasets. To overcome this challenge, a common workaround is
to compute these distances on minibatches i.e. to average the outcome of
several smaller optimal transport problems. We propose in this paper an
extended analysis of this practice, which effects were previously studied in
restricted cases. We first consider a large variety of Optimal Transport
kernels. We notably argue that the minibatch strategy comes with appealing
properties such as unbiased estimators, gradients and a concentration bound
around the expectation, but also with limits: the minibatch OT is not a
distance. To recover some of the lost distance axioms, we introduce a debiased
minibatch OT function and study its statistical and optimisation properties.
Along with this theoretical analysis, we also conduct empirical experiments on
gradient flows, generative adversarial networks (GANs) or color transfer that
highlight the practical interest of this strategy.
| stat |
Bayesian structure learning and sampling of Bayesian networks with the R
package BiDAG | The R package BiDAG implements Markov chain Monte Carlo (MCMC) methods for
structure learning and sampling of Bayesian networks. The package includes
tools to search for a maximum a posteriori (MAP) graph and to sample graphs
from the posterior distribution given the data. A new hybrid approach to
structure learning enables inference in large graphs. In the first step, we
define a reduced search space by means of the PC algorithm or based on prior
knowledge. In the second step, an iterative order MCMC scheme proceeds to
optimize within the restricted search space and estimate the MAP graph.
Sampling from the posterior distribution is implemented using either order or
partition MCMC. The models and algorithms can handle both discrete and
continuous data. The BiDAG package also provides an implementation of MCMC
schemes for structure learning and sampling of dynamic Bayesian networks.
| stat |
Wasserstein distance estimates for the distributions of numerical
approximations to ergodic stochastic differential equations | We present a framework that allows for the non-asymptotic study of the
$2$-Wasserstein distance between the invariant distribution of an ergodic
stochastic differential equation and the distribution of its numerical
approximation in the strongly log-concave case. This allows us to study in a
unified way a number of different integrators proposed in the literature for
the overdamped and underdamped Langevin dynamics. In addition, we analyse a
novel splitting method for the underdamped Langevin dynamics which only
requires one gradient evaluation per time step. Under an additional smoothness
assumption on a $d$--dimensional strongly log-concave distribution with
condition number $\kappa$, the algorithm is shown to produce with an
$\mathcal{O}\big(\kappa^{5/4} d^{1/4}\epsilon^{-1/2} \big)$ complexity samples
from a distribution that, in Wasserstein distance, is at most $\epsilon>0$ away
from the target distribution.
| stat |
Testing nonparametric shape restrictions | We describe and examine a test for a general class of shape constraints, such
as constraints on the signs of derivatives, U-(S-)shape, symmetry,
quasi-convexity, log-convexity, $r$-convexity, among others, in a nonparametric
framework using partial sums empirical processes. We show that, after a
suitable transformation, its asymptotic distribution is a functional of the
standard Brownian motion, so that critical values are available. However, due
to the possible poor approximation of the asymptotic critical values to the
finite sample ones, we also describe a valid bootstrap algorithm.
| stat |
Anomaly detection with superexperts under delayed feedback | The increasing connectivity of data and cyber-physical systems has resulted
in a growing number of cyber-attacks. Real-time detection of such attacks,
through the identification of anomalous activity, is required so that
mitigation and contingent actions can be effectively and rapidly deployed. We
propose a new approach for aggregating unsupervised anomaly detection
algorithms and incorporating feedback when it becomes available. We apply this
approach to open-source real datasets and show that both aggregating models,
which we call experts, and incorporating feedback significantly improve the
performance. An important property of the proposed approaches is their
theoretical guarantees that they perform close to the best superexpert, which
can switch between the best performing experts, in terms of the cumulative
average losses.
| stat |
Winning the Big Data Technologies Horizon Prize: Fast and reliable
forecasting of electricity grid traffic by identification of recurrent
fluctuations | This paper provides a description of the approach and methodology I used in
winning the European Union Big Data Technologies Horizon Prize on data-driven
prediction of electricity grid traffic. The methodology relies on identifying
typical short-term recurrent fluctuations, which is subsequently refined
through a regression-of-fluctuations approach. The key points and strategic
considerations that led to selecting or discarding different methodological
aspects are also discussed. The criteria include adaptability to changing
conditions, reliability with outliers and missing data, robustness to noise,
and efficiency in implementation.
| stat |
Deep ReLU Networks Preserve Expected Length | Assessing the complexity of functions computed by a neural network helps us
understand how the network will learn and generalize. One natural measure of
complexity is how the network distorts length -- if the network takes a
unit-length curve as input, what is the length of the resulting curve of
outputs? It has been widely believed that this length grows exponentially in
network depth. We prove that in fact this is not the case: the expected length
distortion does not grow with depth, and indeed shrinks slightly, for ReLU
networks with standard random initialization. We also generalize this result by
proving upper bounds both for higher moments of the length distortion and for
the distortion of higher-dimensional volumes. These theoretical results are
corroborated by our experiments, which indicate that length distortion remains
modest even after training.
| stat |
Sensitivity analysis based dimension reduction of multiscale models | In this paper, the sensitivity analysis of a single scale model is employed
in order to reduce the input dimensionality of the related multiscale model, in
this way, improving the efficiency of its uncertainty estimation. The approach
is illustrated with two examples: a reaction model and the standard
Ornstein-Uhlenbeck process. Additionally, a counterexample shows that an
uncertain input should not be excluded from uncertainty quantification without
estimating the response sensitivity to this parameter. In particular, an
analysis of the function defining the relation between single scale components
is required to understand whether single scale sensitivity analysis can be used
to reduce the dimensionality of the overall multiscale model input space.
| stat |
Implicit Regularization of Random Feature Models | Random Feature (RF) models are used as efficient parametric approximations of
kernel methods. We investigate, by means of random matrix theory, the
connection between Gaussian RF models and Kernel Ridge Regression (KRR). For a
Gaussian RF model with $P$ features, $N$ data points, and a ridge $\lambda$, we
show that the average (i.e. expected) RF predictor is close to a KRR predictor
with an effective ridge $\tilde{\lambda}$. We show that $\tilde{\lambda} >
\lambda$ and $\tilde{\lambda} \searrow \lambda$ monotonically as $P$ grows,
thus revealing the implicit regularization effect of finite RF sampling. We
then compare the risk (i.e. test error) of the $\tilde{\lambda}$-KRR predictor
with the average risk of the $\lambda$-RF predictor and obtain a precise and
explicit bound on their difference. Finally, we empirically find an extremely
good agreement between the test errors of the average $\lambda$-RF predictor
and $\tilde{\lambda}$-KRR predictor.
| stat |
Subsets and Splits
No saved queries yet
Save your SQL queries to embed, download, and access them later. Queries will appear here once saved.