title
stringlengths 6
244
| abstract
stringlengths 19
5.09k
| label
stringclasses 10
values |
|---|---|---|
Batch-sequential design and heteroskedastic surrogate modeling for delta
smelt conservation | Delta smelt is an endangered fish species in the San Francisco estuary that
have shown an overall population decline over the past 30 years. Researchers
have developed a stochastic, agent-based simulator to virtualize the system,
with the goal of understanding the relative contribution of natural and
anthropogenic factors suggested as playing a role in their decline. However,
the input configuration space is high-dimensional, running the simulator is
time-consuming, and its noisy outputs change nonlinearly in both mean and
variance. Getting enough runs to effectively learn input--output dynamics
requires both a nimble modeling strategy and parallel supercomputer evaluation.
Recent advances in heteroskedastic Gaussian process (HetGP) surrogate modeling
helps, but little is known about how to appropriately plan experiments for
highly distributed simulator evaluation. We propose a batch sequential design
scheme, generalizing one-at-a-time variance-based active learning for HetGP
surrogates, as a means of keeping multi-core cluster nodes fully engaged with
expensive runs. Our acquisition strategy is carefully engineered to favor
selection of replicates which boost statistical and computational efficiencies
when training surrogates to isolate signal in high noise regions. Design and
modeling performance is illustrated on a range of toy examples before embarking
on a large-scale smelt simulation campaign and downstream high-fidelity input
sensitivity analysis.
| stat |
BaySize: Bayesian Sample Size Planning for Phase I Dose-Finding Trials | We propose BaySize, a sample size calculator for phase I clinical trials
using Bayesian models. BaySize applies the concept of effect size in dose
finding, assuming the MTD is defined based on an equivalence interval.
Leveraging a decision framework that involves composite hypotheses, BaySize
utilizes two prior distributions, the fitting prior (for model fitting) and
sampling prior (for data generation), to conduct sample size calculation under
desirable statistical power. Look-up tables are generated to facilitate
practical applications. To our knowledge, BaySize is the first sample size tool
that can be applied to a broad range of phase I trial designs.
| stat |
Fair-by-design explainable models for prediction of recidivism | Recidivism prediction provides decision makers with an assessment of the
likelihood that a criminal defendant will reoffend that can be used in
pre-trial decision-making. It can also be used for prediction of locations
where crimes most occur, profiles that are more likely to commit violent
crimes. While such instruments are gaining increasing popularity, their use is
controversial as they may present potential discriminatory bias in the risk
assessment. In this paper we propose a new fair-by-design approach to predict
recidivism. It is prototype-based, learns locally and extracts empirically the
data distribution. The results show that the proposed method is able to reduce
the bias and provide human interpretable rules to assist specialists in the
explanation of the given results.
| stat |
Particle gradient descent model for point process generation | This paper introduces a generative model for planar point processes in a
square window, built upon a single realization of a stationary, ergodic point
process observed in this window. Inspired by recent advances in gradient
descent methods for maximum entropy models, we propose a method to generate
similar point patterns by jointly moving particles of an initial Poisson
configuration towards a target counting measure. The target measure is
generated via a deterministic gradient descent algorithm, so as to match a set
of statistics of the given, observed realization. Our statistics are estimators
of the multi-scale wavelet phase harmonic covariance, recently proposed in
image modeling. They allow one to capture geometric structures through
multi-scale interactions between wavelet coefficients. Both our statistics and
the gradient descent algorithm scale better with the number of observed points
than the classical k-nearest neighbour distances previously used in generative
models for point processes, based on the rejection sampling or
simulated-annealing. The overall quality of our model is evaluated on point
processes with various geometric structures through spectral and topological
data analysis.
| stat |
Yield curve and macroeconomy interaction: evidence from the
non-parametric functional lagged regression approach | Viewing a yield curve as a sparse collection of measurements on a latent
continuous random function allows us to model it statistically as a sparsely
observed functional time series. Doing so, we use the state-of-the-art methods
in non-parametric statistical inference for sparsely observed functional time
series to analyse the lagged regression dependence of the US Treasury yield
curve on US macroeconomic variables. Our non-parametric analysis confirms
previous findings established under parametric assumptions, namely a strong
impact of the federal funds rate on the short end of the yield curve and a
moderate effect of the annual inflation on the longer end of the yield curve.
| stat |
On systems of quotas based on bankruptcy with a priori unions:
estimating random arrival-style rules | This paper addresses a sampling procedure for estimating extensions of the
random arrival rule to those bankruptcy situations where there exist a priori
unions. It is based on simple random sampling with replacement and it adapts an
estimation method of the Owen value for transferable utility games with a
priori unions, especially useful when the set of involved agents is
sufficiently large. We analyse the theoretical statistical properties of the
resulting estimator as well as we provide some bounds for the incurred error.
Its performance is evaluated on two well-studied examples in literature where
this allocation rule can be exactly obtained. Finally, we apply this sampling
method to provide a new quota system for the milk market in Galicia (Spain) to
check the role of different territorial structures when they are taken as a
priori unions. The resulting quotas estimator is also compared with two
classical rules in bankruptcy literature.
| stat |
A Dynamic Process Reference Model for Sparse Networks with Reciprocity | Many social and other networks exhibit stable size scaling relationships,
such that features such as mean degree or reciprocation rates change slowly or
are approximately constant as the number of vertices increases. Statistical
network models built on top of simple Bernoulli baseline (or reference)
measures often behave unrealistically in this respect, leading to the
development of sparse reference models that preserve features such as mean
degree scaling. In this paper, we generalize recent work on the
micro-foundations of such reference models to the case of sparse directed
graphs with non-vanishing reciprocity, providing a dynamic process
interpretation of the emergence of stable macroscopic behavior.
| stat |
Quantifying the causal effect of speed cameras on road traffic accidents
via an approximate Bayesian doubly robust estimator | This paper quantifies the effect of speed cameras on road traffic collisions
using an approximate Bayesian doubly-robust (DR) causal inference estimation
method. Previous empirical work on this topic, which shows a diverse range of
estimated effects, is based largely on outcome regression (OR) models using the
Empirical Bayes approach or on simple before and after comparisons. Issues of
causality and confounding have received little formal attention. A causal DR
approach combines propensity score (PS) and OR models to give an average
treatment effect (ATE) estimator that is consistent and asymptotically normal
under correct specification of either of the two component models. We develop
this approach within a novel approximate Bayesian framework to derive posterior
predictive distributions for the ATE of speed cameras on road traffic
collisions. Our results for England indicate significant reductions in the
number of collisions at speed cameras sites (mean ATE = -15%). Our proposed
method offers a promising approach for evaluation of transport safety
interventions.
| stat |
Clarifying causal mediation analysis: From simple to more robust
strategies for estimation of marginal natural (in)direct effects | This paper aims to contribute to helping practitioners of causal mediation
analysis gain a better understanding of estimation options. We take as inputs
two familiar strategies (weighting and model-based prediction) and a simple way
of combining them (weighted models), and show how we can generate a range of
estimators with different modeling requirements and robustness properties. The
primary goal is to help build intuitive appreciation for robust estimation that
is conducive to sound practice that does not require advanced statistical
knowledge. A second goal is to provide a "menu" of estimators that
practitioners can choose from for the estimation of marginal natural (in)direct
effects. The estimators generated from this exercise include some that coincide
or are similar to existing estimators and others that have not appeared in the
literature. We use a random continuous weights bootstrap to obtain confidence
intervals, and also derive general asymptotic (sandwich) variance formulas for
the estimators. The estimators are illustrated using data from an adolescent
alcohol use prevention study.
| stat |
Simulation of land use dynamics in Paragominas-PA: differences in
spatial rules between smallholdings and agribusiness areas | The aim of this paper is to present the results of the land use dynamic
simulations in the municipality of Paragominas-PA. The simulation is based on
models built from past land use data and spatial variables of the natural
environment and infrastructure. Two spatial units were analyzed: the central
area of commercial agricultural production and the area of settlements and
smallholdings the east. The results show distinct spatial dynamics between the
analyzed areas, among which we highlight the role of soil characteristics and,
associated with the topography and the occupation history, are part of the
context in which is defined rationality producers. Considering the transition
from forest to pasture in the commercial farming area most often associated
soils are sandy. This raises the following hypothesis: the deforestation that
occurred in the period are related to livestock activities. Livestock favors
access to water and low fertility sands does not affect production. On the
other hand, the soybean expansion occurred preferentially on existing pastures
and on clay soils (Belterra clay), reducing the availability of pastures on
these soils. The relative importance of types of soil increases with time. In
the area of settlements, the transition from forest to pasture and family crops
occurred preferentially on the variegated clay. However, it is possible that
the prevalence of this transition on this texture has been given due more to
the history of occupation of this area. Historically the sandy valleys were the
first to be occupied, and the continuity of the deforestation occurred toward
the slopes dominated by variegated clay and plateaus with Belterra clay. These
associations observed indicate that, within a wider context of social, economic
and political factors, natural variable factors in space are important for the
choice of managements in the properties, but they are done differently in the
territory, and the best knowledge of these relationships are useful for
territorial planning.
| stat |
A computational validation for nonparametric assessment of spatial
trends | The analysis of continuously spatially varying processes usually considers
two sources of variation, namely, the large-scale variation collected by the
trend of the process, and the small-scale variation. Parametric trend models on
latitude and longitude are easy to fit and to interpret. However, the use of
simple parametric models for characterizing spatially varying processes may
lead to misspecification problems if the model is not appropriate. Recently,
Meil\'an-Vila et al. (2019) proposed a goodness-of-fit test based on an
L2-distance for assessing a parametric trend model with correlated errors,
under random design, comparing a parametric and a nonparametric trend
estimators. The present work aims to provide a detailed computational analysis
of the behavior of this approach using different bootstrap algorithms for
calibration, under a fixed-design geostatistical framework. Asymptotic results
for the test are provided and an extensive simulation study, considering
complexities that usually arise in geostatistics, is carried out to illustrate
the performance of the proposal.
| stat |
Centered Partition Process: Informative Priors for Clustering | There is a very rich literature proposing Bayesian approaches for clustering
starting with a prior probability distribution on partitions. Most approaches
assume exchangeability, leading to simple representations in terms of
Exchangeable Partition Probability Functions (EPPF). Gibbs-type priors
encompass a broad class of such cases, including Dirichlet and Pitman-Yor
processes. Even though there have been some proposals to relax the
exchangeability assumption, allowing covariate-dependence and partial
exchangeability, limited consideration has been given on how to include
concrete prior knowledge on the partition. For example, we are motivated by an
epidemiological application, in which we wish to cluster birth defects into
groups and we have prior knowledge of an initial clustering provided by
experts. As a general approach for including such prior knowledge, we propose a
Centered Partition (CP) process that modifies the EPPF to favor partitions
close to an initial one. Some properties of the CP prior are described, a
general algorithm for posterior computation is developed, and we illustrate the
methodology through simulation examples and an application to the motivating
epidemiology study of birth defects.
| stat |
Monotonic Gaussian Process Flow | We propose a new framework for imposing monotonicity constraints in a
Bayesian nonparametric setting based on numerical solutions of stochastic
differential equations. We derive a nonparametric model of monotonic functions
that allows for interpretable priors and principled quantification of
hierarchical uncertainty. We demonstrate the efficacy of the proposed model by
providing competitive results to other probabilistic monotonic models on a
number of benchmark functions. In addition, we consider the utility of a
monotonic random process as a part of a hierarchical probabilistic model; we
examine the task of temporal alignment of time-series data where it is
beneficial to use a monotonic random process in order to preserve the
uncertainty in the temporal warpings.
| stat |
Bayesian neural networks and dimensionality reduction | In conducting non-linear dimensionality reduction and feature learning, it is
common to suppose that the data lie near a lower-dimensional manifold. A class
of model-based approaches for such problems includes latent variables in an
unknown non-linear regression function; this includes Gaussian process latent
variable models and variational auto-encoders (VAEs) as special cases. VAEs are
artificial neural networks (ANNs) that employ approximations to make
computation tractable; however, current implementations lack adequate
uncertainty quantification in estimating the parameters, predictive densities,
and lower-dimensional subspace, and can be unstable and lack interpretability
in practice. We attempt to solve these problems by deploying Markov chain Monte
Carlo sampling algorithms (MCMC) for Bayesian inference in ANN models with
latent variables. We address issues of identifiability by imposing constraints
on the ANN parameters as well as by using anchor points. This is demonstrated
on simulated and real data examples. We find that current MCMC sampling schemes
face fundamental challenges in neural networks involving latent variables,
motivating new research directions.
| stat |
Adaptive Latent Space Tuning for Non-Stationary Distributions | Powerful deep learning tools, such as convolutional neural networks (CNN),
are able to learn the input-output relationships of large complicated systems
directly from data. Encoder-decoder deep CNNs are able to extract features
directly from images, mix them with scalar inputs within a general
low-dimensional latent space, and then generate new complex 2D outputs which
represent complex physical phenomenon. One important challenge faced by deep
learning methods is large non-stationary systems whose characteristics change
quickly with time for which re-training is not feasible. In this paper we
present a method for adaptive tuning of the low-dimensional latent space of
deep encoder-decoder style CNNs based on real-time feedback to quickly
compensate for unknown and fast distribution shifts. We demonstrate our
approach for predicting the properties of a time-varying charged particle beam
in a particle accelerator whose components (accelerating electric fields and
focusing magnetic fields) are also quickly changing with time.
| stat |
Approximating Gaussian Process Emulators with Linear Inequality
Constraints and Noisy Observations via MC and MCMC | Adding inequality constraints (e.g. boundedness, monotonicity, convexity)
into Gaussian processes (GPs) can lead to more realistic stochastic emulators.
Due to the truncated Gaussianity of the posterior, its distribution has to be
approximated. In this work, we consider Monte Carlo (MC) and Markov Chain Monte
Carlo (MCMC) methods. However, strictly interpolating the observations may
entail expensive computations due to highly restrictive sample spaces.
Furthermore, having (constrained) GP emulators when data are actually noisy is
also of interest for real-world implementations. Hence, we introduce a noise
term for the relaxation of the interpolation conditions, and we develop the
corresponding approximation of GP emulators under linear inequality
constraints. We show with various toy examples that the performance of MC and
MCMC samplers improves when considering noisy observations. Finally, on 2D and
5D coastal flooding applications, we show that more flexible and realistic GP
implementations can be obtained by considering noise effects and by enforcing
the (linear) inequality constraints.
| stat |
A robust specification test in linear panel data models | The presence of outlying observations may adversely affect statistical
testing procedures that result in unstable test statistics and unreliable
inferences depending on the distortion in parameter estimates. In spite of the
fact that the adverse effects of outliers in panel data models, there are only
a few robust testing procedures available for model specification. In this
paper, a new weighted likelihood based robust specification test is proposed to
determine the appropriate approach in panel data including individual-specific
components. The proposed test has been shown to have the same asymptotic
distribution as that of most commonly used Hausman's specification test under
null hypothesis of random effects specification. The finite sample properties
of the robust testing procedure are illustrated by means of Monte Carlo
simulations and an economic-growth data from the member countries of the
Organisation for Economic Co-operation and Development. Our records reveal that
the robust specification test exhibit improved performance in terms of size and
power of the test in the presence of contamination.
| stat |
Analyzing Commodity Futures Using Factor State-Space Models with Wishart
Stochastic Volatility | We propose a factor state-space approach with stochastic volatility to model
and forecast the term structure of future contracts on commodities. Our
approach builds upon the dynamic 3-factor Nelson-Siegel model and its 4-factor
Svensson extension and assumes for the latent level, slope and curvature
factors a Gaussian vector autoregression with a multivariate Wishart stochastic
volatility process. Exploiting the conjugacy of the Wishart and the Gaussian
distribution, we develop a computationally fast and easy to implement MCMC
algorithm for the Bayesian posterior analysis. An empirical application to
daily prices for contracts on crude oil with stipulated delivery dates ranging
from one to 24 months ahead show that the estimated 4-factor Svensson model
with two curvature factors provides a good parsimonious representation of the
serial correlation in the individual prices and their volatility. It also shows
that this model has a good out-of-sample forecast performance.
| stat |
A fast and efficient Modal EM algorithm for Gaussian mixtures | In the modal approach to clustering, clusters are defined as the local maxima
of the underlying probability density function, where the latter can be
estimated either non-parametrically or using finite mixture models. Thus,
clusters are closely related to certain regions around the density modes, and
every cluster corresponds to a bump of the density. The Modal EM algorithm is
an iterative procedure that can identify the local maxima of any density
function. In this contribution, we propose a fast and efficient Modal EM
algorithm to be used when the density function is estimated through a finite
mixture of Gaussian distributions with parsimonious component-covariance
structures. After describing the procedure, we apply the proposed Modal EM
algorithm on both simulated and real data examples, showing its high
flexibility in several contexts.
| stat |
A hierarchical model of non-homogeneous Poisson processes for Twitter
retweets | We present a hierarchical model of non-homogeneous Poisson processes (NHPP)
for information diffusion on online social media, in particular Twitter
retweets. The retweets of each original tweet are modelled by a NHPP, for which
the intensity function is a product of time-decaying components and another
component that depends on the follower count of the original tweet author. The
latter allows us to explain or predict the ultimate retweet count by a network
centrality-related covariate. The inference algorithm enables the Bayes factor
to be computed, in order to facilitate model selection. Finally, the model is
applied to the retweet data sets of two hashtags.
| stat |
A New Learning Paradigm for Random Vector Functional-Link Network: RVFL+ | In school, a teacher plays an important role in various classroom teaching
patterns. Likewise to this human learning activity, the learning using
privileged information (LUPI) paradigm provides additional information
generated by the teacher to 'teach' learning models during the training stage.
Therefore, this novel learning paradigm is a typical Teacher-Student
Interaction mechanism. This paper is the first to present a random vector
functional link network based on the LUPI paradigm, called RVFL+. Rather than
simply combining two existing approaches, the newly-derived RVFL+ fills the gap
between classical randomized neural networks and the newfashioned LUPI
paradigm, which offers an alternative way to train RVFL networks. Moreover, the
proposed RVFL+ can perform in conjunction with the kernel trick for highly
complicated nonlinear feature learning, which is termed KRVFL+. Furthermore,
the statistical property of the proposed RVFL+ is investigated, and we present
a sharp and high-quality generalization error bound based on the Rademacher
complexity. Competitive experimental results on 14 real-world datasets
illustrate the great effectiveness and efficiency of the novel RVFL+ and
KRVFL+, which can achieve better generalization performance than
state-of-the-art methods.
| stat |
A Generative Model for Molecular Distance Geometry | Great computational effort is invested in generating equilibrium states for
molecular systems using, for example, Markov chain Monte Carlo. We present a
probabilistic model that generates statistically independent samples for
molecules from their graph representations. Our model learns a low-dimensional
manifold that preserves the geometry of local atomic neighborhoods through a
principled learning representation that is based on Euclidean distance
geometry. In a new benchmark for molecular conformation generation, we show
experimentally that our generative model achieves state-of-the-art accuracy.
Finally, we show how to use our model as a proposal distribution in an
importance sampling scheme to compute molecular properties.
| stat |
An exploratory assessment of a multidimensional healthcare and economic
data on COVID-19 in Nigeria | The coronavirus disease of 2019 (COVID-19) is a pandemic that is ravaging
Nigeria and the world at large. This data article provides a dataset of daily
updates of COVID-19 as reported online by the Nigeria Centre for Disease
Control (NCDC) from February 27, 2020 to September 29, 2020. The data were
obtained through web scraping from different sources and it includes some
economic variables such as the Nigeria budget for each state in 2020,
population estimate, healthcare facilities, and the COVID-19 laboratories in
Nigeria. The dataset has been processed using the standard of the FAIR data
principle which encourages its findability, accessibility, interoperability,
and reusability and will be relevant to researchers in different fields such as
Data Science, Epidemiology, Earth Modelling, and Health Informatics.
| stat |
False discovery rate envelope for functional test statistics | False discovery rate (FDR) is a common way to control the number of false
discoveries in multiple testing. In this paper, the focus is on functional test
statistics which are discretized into $m$ highly correlated hypotheses and thus
resampling based methods are investigated. The aim is to find a graphical
envelope that detects the outcomes of all individual hypotheses by a simple
rule: the hypothesis is rejected if and only if the empirical test statistic is
outside of the envelope. Such an envelope offers a straightforward
interpretation of the test results similarly as in global envelope testing
recently developed for controlling the family-wise error rate. Two different
algorithms are developed to fulfill this aim. The proposed algorithms are
adaptive single threshold procedures which include the estimation of the true
null hypotheses. The new methods are illustrated by two real data examples.
| stat |
Fast calculation of Gaussian Process multiple-fold cross-validation
residuals and their covariances | We generalize fast Gaussian process leave-one-out formulae to multiple-fold
cross-validation, highlighting in turn in broad settings the covariance
structure of cross-validation residuals. The employed approach, that relies on
block matrix inversion via Schur complements, is applied to both Simple and
Universal Kriging frameworks. We illustrate how resulting covariances affect
model diagnostics and how to properly transform residuals in the first place.
Beyond that, we examine how accounting for dependency between such residuals
affect cross-validation-based estimation of the scale parameter. It is found in
two distinct cases, namely in scale estimation and in broader covariance
parameter estimation via pseudo-likelihood, that correcting for covariances
between cross-validation residuals leads back to maximum likelihood estimation
or to an original variation thereof. The proposed fast calculation of Gaussian
Process multiple-fold cross-validation residuals is implemented and benchmarked
against a naive implementation, all in R language. Numerical experiments
highlight the accuracy of our approach as well as the substantial speed-ups
that it enables. It is noticeable however, as supported by a discussion on the
main drivers of computational costs and by a dedicated numerical benchmark,
that speed-ups steeply decline as the number of folds (say, all sharing the
same size) decreases. Overall, our results enable fast multiple-fold
cross-validation, have direct consequences in GP model diagnostics, and pave
the way to future work on hyperparameter fitting as well as on the promising
field of goal-oriented fold design.
| stat |
Random graphs with node and block effects: models, goodness-of-fit
tests, and applications to biological networks | Many popular models from the networks literature can be viewed through a
common lens. We describe it here and call the class of models log-linear ERGMs.
It includes degree-based models, stochastic blockmodels, and combinations of
these. Given the interest in combined node and block effects in network
formation mechanisms, we introduce a general directed relative of the
degree-corrected stochastic blockmodel: an exponential family model we call
$p_1$-SBM. It is a generalization of several well-known variants of the
blockmodel.
We study the problem of testing model fit for the log-linear ERGM class.
The model fitting approach we take, through the use of quick estimation
algorithms borrowed from the contingency table literature and effective
sampling methods rooted in graph theory and algebraic statistics, results in an
exact test whose $p$-value can be approximated efficiently in networks of
moderate sizes.
We showcase the performance of the method on two data sets from biology: the
connectome of \emph{C. elegans} and the interactome of \emph{Arabidopsis
thaliana}. These two networks, a neuronal network and a protein-protein
interaction network, have been popular examples in the network science
literature, but a model-based approach to studying them has been missing thus
far.
| stat |
On Convergence Rate of the Gaussian Belief Propagation Algorithm for
Markov Networks | Gaussian Belief Propagation (BP) algorithm is one of the most important
distributed algorithms in signal processing and statistical learning involving
Markov networks. It is well known that the algorithm correctly computes
marginal density functions from a high dimensional joint density function over
a Markov network in a finite number of iterations when the underlying Gaussian
graph is acyclic. It is also known more recently that the algorithm produces
correct marginal means asymptotically for cyclic Gaussian graphs under the
condition of walk summability. This paper extends this convergence result
further by showing that the convergence is exponential under the walk
summability condition, and provides a simple bound for the convergence rate.
| stat |
Discrete versus continuous domain models for disease mapping | The main goal of disease mapping is to estimate disease risk and identify
high-risk areas. Such analyses are hampered by the limited geographical
resolution of the available data. Typically the available data are counts per
spatial unit and the common approach is the Besag--York--Molli{\'e} (BYM)
model. When precise geocodes are available, it is more natural to use
Log-Gaussian Cox processes (LGCPs). In a simulation study mimicking childhood
leukaemia incidence using actual residential locations of all children in the
canton of Z\"urich, Switzerland, we compare the ability of these models to
recover risk surfaces and identify high-risk areas. We then apply both
approaches to actual data on childhood leukaemia incidence in the canton of
Z\"urich during 1985-2015. We found that LGCPs outperform BYM models in almost
all scenarios considered. Our findings suggest that there are important gains
to be made from the use of LGCPs in spatial epidemiology.
| stat |
Statistical Testing on ASR Performance via Blockwise Bootstrap | A common question being raised in automatic speech recognition (ASR)
evaluations is how reliable is an observed word error rate (WER) improvement
comparing two ASR systems, where statistical hypothesis testing and confidence
interval (CI) can be utilized to tell whether this improvement is real or only
due to random chance. The bootstrap resampling method has been popular for such
significance analysis which is intuitive and easy to use. However, this method
fails in dealing with dependent data, which is prevalent in speech world - for
example, ASR performance on utterances from the same speaker could be
correlated. In this paper we present blockwise bootstrap approach - by dividing
evaluation utterances into nonoverlapping blocks, this method resamples these
blocks instead of original data. We show that the resulting variance estimator
of absolute WER difference between two ASR systems is consistent under mild
conditions. We also demonstrate the validity of blockwise bootstrap method on
both synthetic and real-world speech data.
| stat |
Spatial Resolution Enhancement of Oversampled Images Using Regression
Decomposition and Synthesis | A new statistical model designed for regression analysis with a sparse design
matrix is proposed. This new model utilizes the positions of the limited
non-zero elements in the design matrix to decompose the regression model into
sub-regression models. Statistical inferences are further made on the values of
these limited non-zero elements to provide a reference for synthesizing these
sub-regression models. With this concept of the regression decomposition and
synthesis, the information on the structure of the design matrix can be
incorporated into the regression analysis to provide a more reliable
estimation. The proposed model is then applied to resolve the spatial
resolution enhancement problem for spatially oversampled images. To
systematically evaluate the performance of the proposed model in enhancing the
spatial resolution, the proposed approach is applied to the oversampled images
that are reproduced via random field simulations. These application results
based on different generated scenarios then conclude the effectiveness and the
feasibility of the proposed approach in enhancing the spatial resolution of
spatially oversampled images.
| stat |
Fast and Powerful Conditional Randomization Testing via Distillation | We consider the problem of conditional independence testing, where given a
response $Y$ and covariates $(X,Z)$, we test the null hypothesis that $Y \perp
X \mid Z$. The conditional randomization test (CRT) was recently proposed as a
way to use distributional information about $X\mid Z$ to exactly
(non-asymptotically) control Type-I error using $any$ test statistic in $any$
dimensionality without assuming anything about $Y\mid (X,Z)$. This flexibility
in principle allows one to derive powerful test statistics from complex
state-of-the-art machine learning algorithms while maintaining statistical
validity. Yet the direct use of such advanced test statistics in the CRT is
prohibitively computationally expensive, especially with multiple testing, due
to the CRT's requirement to recompute the test statistic many times on
resampled data. We propose the $distilled~CRT$, a novel approach to using
state-of-the-art machine learning algorithms in the CRT while drastically
reducing the number of times those algorithms need to be run, thereby taking
advantage of their power and the CRT's statistical guarantees without suffering
the usual computational expense. In addition to distillation, we propose a
number of other tricks like screening and recycling computations to further
speed up the CRT without sacrificing its high power and exact validity. Indeed,
we show in simulations that all our proposals combined lead to a test that has
similar power to the CRT but requires orders of magnitude less computation,
making it a practical tool even for large data sets. We demonstrate these
benefits on a breast cancer dataset by identifying biomarkers related to cancer
stage.
| stat |
Robust Clustering with Normal Mixture Models: A Pseudo
$\beta$-Likelihood Approach | As in other estimation scenarios, likelihood based estimation in the normal
mixture set-up is highly non-robust against model misspecification and presence
of outliers (apart from being an ill-posed optimization problem). We propose a
robust alternative to the ordinary likelihood approach for this estimation
problem which performs simultaneous estimation and data clustering and leads to
subsequent anomaly detection. To invoke robustness, we follow, in spirit, the
methodology based on the minimization of the density power divergence (or
alternatively, the maximization of the $\beta$-likelihood) under suitable
constraints. An iteratively reweighted least squares approach has been followed
in order to compute our estimators for the component means (or equivalently
cluster centers) and component dispersion matrices which leads to simultaneous
data clustering. Some exploratory techniques are also suggested for anomaly
detection, a problem of great importance in the domain of statistics and
machine learning. Existence and consistency of the estimators are established
under the aforesaid constraints. We validate our method with simulation studies
under different set-ups; it is seen to perform competitively or better compared
to the popular existing methods like K-means and TCLUST, especially when the
mixture components (i.e., the clusters) share regions with significant overlap
or outlying clusters exist with small but non-negligible weights. Two real
datasets are also used to illustrate the performance of our method in
comparison with others along with an application in image processing. It is
observed that our method detects the clusters with lower misclassification
rates and successfully points out the outlying (anomalous) observations from
these datasets.
| stat |
Bayesian Optimization with Approximate Set Kernels | We propose a practical Bayesian optimization method over sets, to minimize a
black-box function that takes a set as a single input. Because set inputs are
permutation-invariant, traditional Gaussian process-based Bayesian optimization
strategies which assume vector inputs can fall short. To address this, we
develop a Bayesian optimization method with \emph{set kernel} that is used to
build surrogate functions. This kernel accumulates similarity over set elements
to enforce permutation-invariance, but this comes at a greater computational
cost. To reduce this burden, we propose two key components: (i) a more
efficient approximate set kernel which is still positive-definite and is an
unbiased estimator of the true set kernel with upper-bounded variance in terms
of the number of subsamples, (ii) a constrained acquisition function
optimization over sets, which uses symmetry of the feasible region that defines
a set input. Finally, we present several numerical experiments which
demonstrate that our method outperforms other methods.
| stat |
Adversarial Regression. Generative Adversarial Networks for Non-Linear
Regression: Theory and Assessment | Adversarial Regression is a proposition to perform high dimensional
non-linear regression with uncertainty estimation. We used Conditional
Generative Adversarial Network to obtain an estimate of the full predictive
distribution for a new observation. Generative Adversarial Networks (GAN) are
implicit generative models which produce samples from a distribution
approximating the distribution of the data. The conditional version of it
(CGAN) takes the following expression: $\min\limits_G \max\limits_D V(D, G) =
\mathbb{E}_{x\sim p_{r}(x)} [log(D(x, y))] + \mathbb{E}_{z\sim p_{z}(z)} [log
(1-D(G(z, y)))]$. An approximate solution can be found by training
simultaneously two neural networks to model D and G and feeding G with a random
noise vector $z$. After training, we have that $G(z, y)\mathrel{\dot\sim}
p_{data}(x, y)$. By fixing $y$, we have $G(z|y) \mathrel{\dot\sim}
p{data}(x|y)$. By sampling $z$, we can therefore obtain samples following
approximately $p(x|y)$, which is the predictive distribution of $x$ for a new
$y$. We ran experiments to test various loss functions, data distributions,
sample size, size of the noise vector, etc. Even if we observed differences, no
experiment outperformed consistently the others. The quality of CGAN for
regression relies on fine-tuning a range of hyperparameters. In a broader view,
the results show that CGANs are very promising methods to perform uncertainty
estimation for high dimensional non-linear regression.
| stat |
Learning Temporal Evolution of Spatial Dependence with Generalized
Spatiotemporal Gaussian Process Models | A large number of scientific studies and engineering problems involve
high-dimensional spatiotemporal data with complicated relationships. In this
paper, we focus on a type of space-time interaction named \emph{temporal
evolution of spatial dependence (TESD)}, which is a zero time-lag
spatiotemporal covariance. For this purpose, we propose a novel Bayesian
nonparametric method based on non-stationary spatiotemporal Gaussian process
(STGP). The classic STGP has a covariance kernel separable in space and time,
failed to characterize TESD. More recent works on non-separable STGP treat
location and time together as a joint variable, which is unnecessarily
inefficient. We generalize STGP (gSTGP) to introduce the time-dependence to the
spatial kernel by varying its eigenvalues over time in the Mercer's
representation. The resulting non-separable non-stationary covariance model
bares a quasi Kronecker sum structure. Finally, a hierarchical Bayesian model
for the joint covariance is proposed to allow for full flexibility in learning
TESD. A simulation study and a longitudinal neuroimaging analysis on
Alzheimer's patients demonstrate that the proposed methodology is
(statistically) effective and (computationally) efficient in characterizing
TESD. Theoretic properties of gSTGP including posterior contraction (for
covariance) are also studied.
| stat |
Optimal Paired Comparison Experiments for Second-Order Interactions | In real life situations often paired comparisons involving alternatives of
either full or partial profiles to mitigate cognitive burden are presented. For
this situation the problem of finding optimal designs is considered in the
presence of second-order interactions when all attributes have general common
number of levels.
| stat |
Most Likely Optimal Subsampled Markov Chain Monte Carlo | Markov Chain Monte Carlo (MCMC) requires to evaluate the full data likelihood
at different parameter values iteratively and is often computationally
infeasible for large data sets. In this paper, we propose to approximate the
log-likelihood with subsamples taken according to nonuniform subsampling
probabilities, and derive the most likely optimal (MLO) subsampling
probabilities for better approximation. Compared with existing subsampled MCMC
algorithm with equal subsampling probabilities, our MLO subsampled MCMC has a
higher estimation efficiency with the same subsampling ratio. We also derive a
formula using the asymptotic distribution of the subsampled log-likelihood to
determine the required subsample size in each MCMC iteration for a given level
of precision. This formula is used to develop an adaptive version of the MLO
subsampled MCMC algorithm. Numerical experiments demonstrate that the proposed
method outperforms the uniform subsampled MCMC.
| stat |
Efficient augmentation and relaxation learning for individualized
treatment rules using observational data | Individualized treatment rules aim to identify if, when, which, and to whom
treatment should be applied. A globally aging population, rising healthcare
costs, and increased access to patient-level data have created an urgent need
for high-quality estimators of individualized treatment rules that can be
applied to observational data. A recent and promising line of research for
estimating individualized treatment rules recasts the problem of estimating an
optimal treatment rule as a weighted classification problem. We consider a
class of estimators for optimal treatment rules that are analogous to convex
large-margin classifiers. The proposed class applies to observational data and
is doubly-robust in the sense that correct specification of either a propensity
or outcome model leads to consistent estimation of the optimal individualized
treatment rule. Using techniques from semiparametric efficiency theory, we
derive rates of convergence for the proposed estimators and use these rates to
characterize the bias-variance trade-off for estimating individualized
treatment rules with classification-based methods. Simulation experiments
informed by these results demonstrate that it is possible to construct new
estimators within the proposed framework that significantly outperform existing
ones. We illustrate the proposed methods using data from a labor training
program and a study of inflammatory bowel syndrome.
| stat |
Corella: A Private Multi Server Learning Approach based on Correlated
Queries | The emerging applications of machine learning algorithms on mobile devices
motivate us to offload the computation tasks of training a model or deploying a
trained one to the cloud or at the edge of the network. One of the major
challenges in this setup is to guarantee the privacy of the client data.
Various methods have been proposed to protect privacy in the literature. Those
include (i) adding noise to the client data, which reduces the accuracy of the
result, (ii) using secure multiparty computation (MPC), which requires
significant communication among the computing nodes or with the client, (iii)
relying on homomorphic encryption (HE) methods, which significantly increases
computation load at the servers. In this paper, we propose $\textit{Corella}$
as an alternative approach to protect the privacy of data. The proposed scheme
relies on a cluster of servers, where at most $T \in \mathbb{N}$ of them may
collude, each running a learning model (e.g., a deep neural network). Each
server is fed with the client data, added with $\textit{strong}$ noise,
independent from user data. The variance of the noise is set to be large enough
to make the information leakage to any subset of up to $T$ servers
information-theoretically negligible. On the other hand, the added noises for
different servers are $\textit{correlated}$. This correlation among the queries
allows the parameters of the models running on different servers to be
$\textit{trained}$ such that the client can mitigate the contribution of the
noises by combining the outputs of the servers, and recover the final result
with high accuracy and with a minor computational effort. Simulation results
for various datasets demonstrate the accuracy of the proposed approach for the
classification, using deep neural networks, and the autoencoder, as supervised
and unsupervised learning tasks, respectively.
| stat |
A Bayesian response-adaptive dose finding and comparative effectiveness
trial | Aims: Combinations of treatments can offer additional benefit over the
treatments individually. However, trials of these combinations are lower
priority than the development of novel therapies, which can restrict funding,
timelines and patient availability. This paper develops a novel trial design to
facilitate the evaluation of novel combination therapies that combines elements
of phase II and phase III trials. Methods: This trial uses response adaptive
randomisation to increase the information collected about successful novel drug
combinations and Bayesian dose-response modelling to undertake a
comparative-effectiveness analysis for the most successful dose combination
against a relevant comparator. We used simulation methods to evaluate the
probability of selecting the correct optimal dose combination, the operating
characteristics and predictive power of this design for a trial in pain
management and sedation in paediatric emergency departments. Results: With 410
participants, 5 interim updates of the randomisation ratio and a probability of
effectiveness of 0.93, 0.88 and 0.83 for the three dose combinations, we have
an 83% chance of randomising the largest number of patients to the drug with
the highest probability of effectiveness. Based on this adaptive randomisation
procedure, the comparative effectiveness analysis has a type I error of less
than 5% and a 93% chance of correctly concluding non-inferiority when the
probability of effectiveness for the optimal combination therapy is 0.9. In
this case, the trial has a 77% chance of meeting its dual aims of dose finding
and comparative effectiveness. Finally, the Bayesian predictive power of the
trial is over 90%. Conclusion: The proposed trial has high potential to meet
the dual study objectives within a feasible level of recruitment, minimising
the administrative burden and recruitment time for a trial.
| stat |
Modelling multi-scale state-switching functional data with hidden Markov
models | Data sets comprised of sequences of curves sampled at high frequencies in
time are increasingly common in practice, but they can exhibit complicated
dependence structures that cannot be modelled using common methods of
Functional Data Analysis (FDA). We detail a hierarchical approach which treats
the curves as observations from a hidden Markov model (HMM). The distribution
of each curve is then defined by another fine-scale model which may involve
auto-regression and require data transformations using moving-window summary
statistics or Fourier analysis. This approach is broadly applicable to
sequences of curves exhibiting intricate dependence structures. As a case
study, we use this framework to model the fine-scale kinematic movement of a
northern resident killer whale (Orcinus orca) off the coast of British
Columbia, Canada. Through simulations, we show that our model produces more
interpretable state estimation and more accurate parameter estimates compared
to existing methods.
| stat |
On Scalable Particle Markov Chain Monte Carlo | Particle Markov Chain Monte Carlo (PMCMC) is a general approach to carry out
Bayesian inference in non-linear and non-Gaussian state space models. Our
article shows how to scale up PMCMC in terms of the number of observations and
parameters by expressing the target density of the PMCMC in terms of the basic
uniform or standard normal random numbers, instead of the particles, used in
the sequential Monte Carlo algorithm. Parameters that can be drawn efficiently
conditional on the particles are generated by particle Gibbs. All the other
parameters are drawn by conditioning on the basic uniform or standard normal
random variables; e.g. parameters that are highly correlated with the states,
or parameters whose generation is expensive when conditioning on the states.
The performance of this hybrid sampler is investigated empirically by applying
it to univariate and multivariate stochastic volatility models having both a
large number of parameters and a large number of latent states and shows that
it is much more efficient than competing PMCMC methods. We also show that the
proposed hybrid sampler is ergodic.
| stat |
Testing for equality between conditional copulas given discretized
conditioning events | Several procedures have been recently proposed to test the simplifying
assumption for conditional copulas. Instead of considering pointwise
conditioning events, we study the constancy of the conditional dependence
structure when some covariates belong to general borelian conditioning subsets.
Several test statistics based on the equality of conditional Kendall's tau are
introduced, and we derive their asymptotic distributions under the null. When
such conditioning events are not fixed ex ante, we propose a data-driven
procedure to recursively build such relevant subsets. It is based on decision
trees that maximize the differences between the conditional Kendall's taus
corresponding to the leaves of the trees. The performances of such tests are
illustrated in a simulation experiment. Moreover, a study of the conditional
dependence between financial stock returns is managed, given some clustering of
their past values. The last application deals with the conditional dependence
between coverage amounts in an insurance dataset.
| stat |
Distributionally Robust Parametric Maximum Likelihood Estimation | We consider the parameter estimation problem of a probabilistic generative
model prescribed using a natural exponential family of distributions. For this
problem, the typical maximum likelihood estimator usually overfits under
limited training sample size, is sensitive to noise and may perform poorly on
downstream predictive tasks. To mitigate these issues, we propose a
distributionally robust maximum likelihood estimator that minimizes the
worst-case expected log-loss uniformly over a parametric Kullback-Leibler ball
around a parametric nominal distribution. Leveraging the analytical expression
of the Kullback-Leibler divergence between two distributions in the same
natural exponential family, we show that the min-max estimation problem is
tractable in a broad setting, including the robust training of generalized
linear models. Our novel robust estimator also enjoys statistical consistency
and delivers promising empirical results in both regression and classification
tasks.
| stat |
Optimal Change Point Detection and Localization in Sparse Dynamic
Networks | We study the problem of change point localization in dynamic networks models.
We assume that we observe a sequence of independent adjacency matrices of the
same size, each corresponding to a realization of an unknown inhomogeneous
Bernoulli model. The underlying distribution of the adjacency matrices are
piecewise constant, and may change over a subset of the time points, called
change points. We are concerned with recovering the unknown number and
positions of the change points. In our model setting we allow for all the model
parameters to change with the total number of time points, including the
network size, the minimal spacing between consecutive change points, the
magnitude of the smallest change and the degree of sparsity of the networks. We
first identify a region of impossibility in the space of the model parameters
such that no change point estimator is provably consistent if the data are
generated according to parameters falling in that region. We propose a
computationally-simple algorithm for network change point localization, called
Network Binary Segmentation, that relies on weighted averages of the adjacency
matrices. We show that Network Binary Segmentation is consistent over a range
of the model parameters that nearly cover the complement of the impossibility
region, thus demonstrating the existence of a phase transition for the problem
at hand. Next, we devise a more sophisticated algorithm based on singular value
thresholding, called Local Refinement, that delivers more accurate estimates of
the change point locations. Under appropriate conditions, Local Refinement
guarantees a minimax optimal rate for network change point localization while
remaining computationally feasible.
| stat |
Generating Higher-Fidelity Synthetic Datasets with Privacy Guarantees | This paper considers the problem of enhancing user privacy in common machine
learning development tasks, such as data annotation and inspection, by
substituting the real data with samples form a generative adversarial network.
We propose employing Bayesian differential privacy as the means to achieve a
rigorous theoretical guarantee while providing a better privacy-utility
trade-off. We demonstrate experimentally that our approach produces
higher-fidelity samples, compared to prior work, allowing to (1) detect more
subtle data errors and biases, and (2) reduce the need for real data labelling
by achieving high accuracy when training directly on artificial samples.
| stat |
A comparison of learning rate selection methods in generalized Bayesian
inference | Generalized Bayes posterior distributions are formed by putting a fractional
power on the likelihood before combining with the prior via Bayes's formula.
This fractional power, which is often viewed as a remedy for potential model
misspecification bias, is called the learning rate, and a number of data-driven
learning rate selection methods have been proposed in the recent literature.
Each of these proposals has a different focus, a different target they aim to
achieve, which makes them difficult to compare. In this paper, we provide a
direct head-to-head comparison of these learning rate selection methods in
various misspecified model scenarios, in terms of several relevant metrics, in
particular, coverage probability of the generalized Bayes credible regions. In
some examples all the methods perform well, while in others the
misspecification is too severe to be overcome, but we find that the so-called
generalized posterior calibration algorithm tends to outperform the others in
terms of credible region coverage probability.
| stat |
Causal Learning via Manifold Regularization | This paper frames causal structure estimation as a machine learning task. The
idea is to treat indicators of causal relationships between variables as
`labels' and to exploit available data on the variables of interest to provide
features for the labelling task. Background scientific knowledge or any
available interventional data provide labels on some causal relationships and
the remainder are treated as unlabelled. To illustrate the key ideas, we
develop a distance-based approach (based on bivariate histograms) within a
manifold regularization framework. We present empirical results on three
different biological data sets (including examples where causal effects can be
verified by experimental intervention), that together demonstrate the efficacy
and general nature of the approach as well as its simplicity from a user's
point of view.
| stat |
Inference from Non-Random Samples Using Bayesian Machine Learning | We consider inference from non-random samples in data-rich settings where
high-dimensional auxiliary information is available both in the sample and the
target population, with survey inference being a special case. We propose a
regularized prediction approach that predicts the outcomes in the population
using a large number of auxiliary variables such that the ignorability
assumption is reasonable while the Bayesian framework is straightforward for
quantification of uncertainty. Besides the auxiliary variables, inspired by
Little & An (2004), we also extend the approach by estimating the propensity
score for a unit to be included in the sample and also including it as a
predictor in the machine learning models. We show through simulation studies
that the regularized predictions using soft Bayesian additive regression trees
yield valid inference for the population means and coverage rates close to the
nominal levels. We demonstrate the application of the proposed methods using
two different real data applications, one in a survey and one in an
epidemiology study.
| stat |
Hamiltonian Monte Carlo on Symmetric and Homogeneous Spaces via
Symplectic Reduction | The Hamiltonian Monte Carlo method generates samples by introducing a
mechanical system that explores the target density. For distributions on
manifolds it is not always simple to perform the mechanics as a result of the
lack of global coordinates, the constraints of the manifold, and the
requirement to compute the geodesic flow. In this paper we explain how to
construct the Hamiltonian system on naturally reductive homogeneous spaces
using symplectic reduction, which lifts the HMC scheme to a matrix Lie group
with global coordinates and constant metric. This provides a general framework
that is applicable to many manifolds that arise in applications, such as
hyperspheres, hyperbolic spaces, symmetric positive-definite matrices,
Grassmannian, and Stiefel manifolds.
| stat |
Batch Normalization Provably Avoids Rank Collapse for Randomly
Initialised Deep Networks | Randomly initialized neural networks are known to become harder to train with
increasing depth, unless architectural enhancements like residual connections
and batch normalization are used. We here investigate this phenomenon by
revisiting the connection between random initialization in deep networks and
spectral instabilities in products of random matrices. Given the rich
literature on random matrices, it is not surprising to find that the rank of
the intermediate representations in unnormalized networks collapses quickly
with depth. In this work we highlight the fact that batch normalization is an
effective strategy to avoid rank collapse for both linear and ReLU networks.
Leveraging tools from Markov chain theory, we derive a meaningful lower rank
bound in deep linear networks. Empirically, we also demonstrate that this rank
robustness generalizes to ReLU nets. Finally, we conduct an extensive set of
experiments on real-world data sets, which confirm that rank stability is
indeed a crucial condition for training modern-day deep neural architectures.
| stat |
Statistical post-processing of heat index ensemble forecasts: is there a
royal road? | We investigate the effect of statistical post-processing on the probabilistic
skill of discomfort index (DI) and indoor wet-bulb globe temperature (WBGTid)
ensemble forecasts, both calculated from the corresponding forecasts of
temperature and dew point temperature. Two different methodological approaches
to calibration are compared. In the first case, we start with joint
post-processing of the temperature and dew point forecasts and then create
calibrated samples of DI and WBGTid using samples from the obtained bivariate
predictive distributions. This approach is compared with direct post-processing
of the heat index ensemble forecasts. For this purpose, a novel ensemble model
output statistics model based on a generalized extreme value distribution is
proposed. The predictive performance of both methods is tested on the
operational temperature and dew point ensemble forecasts of the European Centre
for Medium-Range Weather Forecasts and the corresponding forecasts of DI and
WBGTid. For short lead times (up to day 6), both approaches significantly
improve the forecast skill. Among the competing post-processing methods, direct
calibration of heat indices exhibits the best predictive performance, very
closely followed by the more general approach based on joint calibration of
temperature and dew point temperature. Additionally, a machine learning
approach is tested and shows comparable performance for the case when one is
interested only in forecasting heat index warning level categories.
| stat |
On the rate of convergence of fully connected very deep neural network
regression estimates | Recent results in nonparametric regression show that deep learning, i.e.,
neural network estimates with many hidden layers, are able to circumvent the
so-called curse of dimensionality in case that suitable restrictions on the
structure of the regression function hold. One key feature of the neural
networks used in these results is that their network architecture has a further
constraint, namely the network sparsity. In this paper we show that we can get
similar results also for least squares estimates based on simple fully
connected neural networks with ReLU activation functions. Here either the
number of neurons per hidden layer is fixed and the number of hidden layers
tends to infinity suitably fast for sample size tending to infinity, or the
number of hidden layers is bounded by some logarithmic factor in the sample
size and the number of neurons per hidden layer tends to infinity suitably fast
for sample size tending to infinity. The proof is based on new approximation
results concerning deep neural networks.
| stat |
Adaptive treatment allocation and selection in multi-arm clinical
trials: a Bayesian perspective | Clinical trials are an instrument for making informed decisions based on
evidence from well-designed experiments. Here we consider adaptive designs
mainly from the perspective of multi-arm Phase II clinical trials, in which one
or more experimental treatments are compared to a control. Treatment allocation
of individual trial participants is assumed to take place according to a fixed
block randomization, albeit with an important twist: The performance of each
treatment arm is assessed after every measured outcome, in terms of the
posterior distribution of a corresponding model parameter. Different treatments
arms are then compared to each other, according to pre-defined criteria and
using the joint posterior as the basis for such assessment. If a treatment is
found to be sufficiently clearly inferior to the currently best candidate, it
can be closed off either temporarily or permanently from further participant
accrual. The latter possibility provides a method for adaptive treatment
selection, including early stopping of the trial. The main development in the
paper is in terms of binary outcomes, but some extensions, notably for handling
time-to-event data, are discussed as well. The presentation is to a large
extent comparative and expository.
| stat |
Combining Cluster Sampling and Link-Tracing Sampling to Estimate Totals
and Means of Hidden Populations in Presence of Heterogeneous Probabilities of
Links | We propose Horvitz-Thompson-like and Hajek-like estimators of the total and
mean of the values of a variable of interest associated with the elements of a
hard-to-reach population sampled by the variant of link-tracing sampling
proposed by Felix-Medina and Thompson (2004). As examples of this type of
population are drug users, homeless people and sex workers. In this sampling
variant, a frame of venues or places where the members of the population tend
to gather, such as parks and bars, is constructed. The frame is not assumed to
cover the whole population. An initial cluster sample of elements is selected
from the frame, where the clusters are the venues, and the elements in the
initial sample are asked to name their contacts who are also members of the
population. The sample size is increased by including in the sample the named
elements who are not in the initial sample. The proposed estimators do not use
design-based inclusion probabilities, but model-based inclusion probabilities
which are derived from a model proposed by Felix-Medina et al. (2015) and are
estimated by maximum likelihood estimators. The inclusion probabilities are
assumed to be heterogeneous, that is, that they depend on the sampled people.
Estimates of the variances of the proposed estimators are obtained by bootstrap
and they are used to construct confidence intervals of the totals and means.
The performance of the proposed estimators and confidence intervals is
evaluated by two numerical studies, one of them based on real data, and the
results show that their performance is acceptable.
| stat |
Effect of COVID-19 on noise pollution change in Dublin, Ireland | Noise pollution is considered to be the third most hazardous pollution after
air and water pollution by the World Health Organization (WHO). Short as well
as long-term exposure to noise pollution has several adverse effects on humans,
ranging from psychiatric disorders such as anxiety and depression,
hypertension, hormonal dysfunction, and blood pressure rise leading to
cardiovascular disease. One of the major sources of noise pollution is road
traffic. The WHO reports that around 40% of Europe's population are currently
exposed to high noise levels. This study investigates noise pollution in
Dublin, Ireland before and after the lockdown imposed as a result of the
COVID-19 pandemic. The analysis was performed using 2020 hourly data from 12
noise monitoring stations. More than 80% of stations recorded high noise levels
for more that 60% of the time before the lockdown in Dublin. However, a
significant reduction in average and minimum noise levels was observed at all
stations during the lockdown period and this can be attributed to reductions in
both road and air traffic movements.
| stat |
Roughness Penalty for liquid Scorecards | A liquid scorecard has liquid characteristics, for which the characteristic
score is a smooth function of the characteristic over a liquid range. The
smooth function is based on B-splines, typically cubic. In contrast, the
characteristic scores for traditional scorecards are step functions of the
characteristics. Previously, there were two ways to control the smoothness of
the liquid characteristic score: (1) coarse classing where the fewer the number
of classes, the smoother the curve; (2) the penalty parameter, which penalizes
the norm of the score coefficient vector. However, in classical cubic spline
fitting theory, a direct measure of curve roughness is used as a penalty term
in the fitting objective function. In this paper, I work out the details of
this concept for our characteristic scores, which are linear functions of a
cubic B-spline basis. The roughness penalty is the integral of the second
derivative squared. As you vary the characteristic smoothness parameter from
zero to infinity, the characteristic score goes from being rough to being very
smooth. As one moves from rough to smooth, the palatable characteristic score
jumps off the page. This is illustrated by a case study. This case study also
shows that smoothness parameters, which maximize validation divergence, do not
always yield the most palatable model.
| stat |
Modeling massive multivariate spatial data with the basis graphical
lasso | We propose a new modeling framework for highly multivariate spatial processes
that synthesizes ideas from recent multiscale and spectral approaches with
graphical models. The basis graphical lasso writes a univariate Gaussian
process as a linear combination of basis functions weighted with entries of a
Gaussian graphical vector whose graph is estimated from optimizing an $\ell_1$
penalized likelihood. This paper extends the setting to a multivariate Gaussian
process where the basis functions are weighted with Gaussian graphical vectors.
We motivate a model where the basis functions represent different levels of
resolution and the graphical vectors for each level are assumed to be
independent. Using an orthogonal basis grants linear complexity and memory
usage in the number of spatial locations, the number of basis functions, and
the number of realizations. An additional fusion penalty encourages a
parsimonious conditional independence structure in the multilevel graphical
model. We illustrate our method on a large climate ensemble from the National
Center for Atmospheric Research's Community Atmosphere Model that involves 40
spatial processes.
| stat |
Estimating the impact of treatment compliance over time on smoking
cessation using data from ecological momentary assessments (EMA) | The Wisconsin Smoker's Health Study (WSHS2) was a longitudinal trial
conducted to compare the effectiveness of two commonly used smoking cessation
treatments, varenicline and combination nicotine replacement therapy (cNRT)
with the less intense standard of care, nicotine patch. The main outcome of the
WSHS2 study was that all three treatments had equivalent treatment effects.
However, in-depth analysis of the compliance data collected via ecological
momentary assessment (EMA) were not analyzed. Compliance to the treatment
regimens may represent a confounder as varenicline and cNRT are more intense
treatments and would likely have larger treatment effects if all subjects
complied. In order to estimate the causal compliance effect, we view the
counterfactual, the outcome that would have been observed if the subject was
allocated to the treatment counter to the fact, as a missing data problem and
proceed to impute the counterfactual. Our contribution to the methodological
literature lies in the extension of this idea to a more general analytic
approach that includes mediators and confounders of the mediator-outcome
relationship. Simulation results suggest that our method works well and
application to the WSHS2 data suggest that the treatment effect of nicotine
patch, varenicline, and cNRT are equivalent after accounting for differences in
treatment compliance.
| stat |
Social Learning and the Accuracy-Risk Trade-off in the Wisdom of the
Crowd | How do we design and deploy crowdsourced prediction platforms for real-world
applications where risk is an important dimension of prediction performance? To
answer this question, we conducted a large online Wisdom of the Crowd study
where participants predicted the prices of real financial assets (e.g. S&P
500). We observe a Pareto frontier between accuracy of prediction and risk, and
find that this trade-off is mediated by social learning i.e. as social learning
is increasingly leveraged, it leads to lower accuracy but also lower risk. We
also observe that social learning leads to superior accuracy during one of our
rounds that occurred during the high market uncertainty of the Brexit vote. Our
results have implications for the design of crowdsourced prediction platforms:
for example, they suggest that the performance of the crowd should be more
comprehensively characterized by using both accuracy and risk (as is standard
in financial and statistical forecasting), in contrast to prior work where risk
of prediction has been overlooked.
| stat |
Bayesian Fusion: Scalable unification of distributed statistical
analyses | There has recently been considerable interest in addressing the problem of
unifying distributed statistical analyses into a single coherent inference.
This problem naturally arises in a number of situations, including in big-data
settings, when working under privacy constraints, and in Bayesian model choice.
The majority of existing approaches have relied upon convenient approximations
of the distributed analyses. Although typically being computationally
efficient, and readily scaling with respect to the number of analyses being
unified, approximate approaches can have significant shortcomings -- the
quality of the inference can degrade rapidly with the number of analyses being
unified, and can be substantially biased even when unifying a small number of
analyses that do not concur. In contrast, the recent Fusion approach of Dai et
al. (2019) is a rejection sampling scheme which is readily parallelisable and
is exact (avoiding any form of approximation other than Monte Carlo error),
albeit limited in applicability to unifying a small number of low-dimensional
analyses. In this paper we introduce a practical Bayesian Fusion approach. We
extend the theory underpinning the Fusion methodology and, by embedding it
within a sequential Monte Carlo algorithm, we are able to recover the correct
target distribution. By means of extensive guidance on the implementation of
the approach, we demonstrate theoretically and empirically that Bayesian Fusion
is robust to increasing numbers of analyses, and coherently unifying analyses
which do not concur. This is achieved while being computationally competitive
with approximate schemes.
| stat |
On the Privacy Risks of Algorithmic Fairness | Algorithmic fairness and privacy are essential pillars of trustworthy machine
learning. Fair machine learning aims at minimizing discrimination against
protected groups by, for example, imposing a constraint on models to equalize
their behavior across different groups. This can subsequently change the
influence of training data points on the fair model, in a disproportionate way.
We study how this can change the information leakage of the model about its
training data. We analyze the privacy risks of group fairness (e.g., equalized
odds) through the lens of membership inference attacks: inferring whether a
data point is used for training a model. We show that fairness comes at the
cost of privacy, and this cost is not distributed equally: the information
leakage of fair models increases significantly on the unprivileged subgroups,
which are the ones for whom we need fair learning. We show that the more biased
the training data is, the higher the privacy cost of achieving fairness for the
unprivileged subgroups will be. We provide comprehensive empirical analysis for
general machine learning algorithms.
| stat |
Direct Estimation of Difference Between Structural Equation Models in
High Dimensions | Discovering cause-effect relationships between variables from observational
data is a fundamental challenge in many scientific disciplines. However, in
many situations it is desirable to directly estimate the change in causal
relationships across two different conditions, e.g., estimating the change in
genetic expression across healthy and diseased subjects can help isolate
genetic factors behind the disease. This paper focuses on the problem of
directly estimating the structural difference between two causal DAGs, having
the same topological ordering, given two sets of samples drawn from the
individual DAGs. We present an algorithm that can recover the difference-DAG in
$O(d \log p)$ samples, where $d$ is related to the number of edges in the
difference-DAG. We also show that any method requires at least $\Omega(d \log
p/d)$ samples to learn difference DAGs with at most $d$ parents per node. We
validate our theoretical results with synthetic experiments and show that our
method out-performs the state-of-the-art.
| stat |
Private Protocols for U-Statistics in the Local Model and Beyond | In this paper, we study the problem of computing $U$-statistics of degree
$2$, i.e., quantities that come in the form of averages over pairs of data
points, in the local model of differential privacy (LDP). The class of
$U$-statistics covers many statistical estimates of interest, including Gini
mean difference, Kendall's tau coefficient and Area under the ROC Curve (AUC),
as well as empirical risk measures for machine learning problems such as
ranking, clustering and metric learning. We first introduce an LDP protocol
based on quantizing the data into bins and applying randomized response, which
guarantees an $\epsilon$-LDP estimate with a Mean Squared Error (MSE) of
$O(1/\sqrt{n}\epsilon)$ under regularity assumptions on the $U$-statistic or
the data distribution. We then propose a specialized protocol for AUC based on
a novel use of hierarchical histograms that achieves MSE of
$O(\alpha^3/n\epsilon^2)$ for arbitrary data distribution. We also show that
2-party secure computation allows to design a protocol with MSE of
$O(1/n\epsilon^2)$, without any assumption on the kernel function or data
distribution and with total communication linear in the number of users $n$.
Finally, we evaluate the performance of our protocols through experiments on
synthetic and real datasets.
| stat |
A copula-based time series model for global horizontal irradiation | The increasing importance of solar power for electricity generation leads to
an increasing demand for probabilistic forecasting of local and aggregated PV
yields. In this paper we use an indirect modeling approach for hourly medium to
long term local PV yields based on publicly available irradiation data. We
suggest a time series model for global horizontal irradiation for which it is
easy to generate an arbitrary number of scenarios and thus allows for
multivariate probabilistic forecasts for arbitrary time horizons. In contrast
to many simplified models that have been considered in the literature so far it
features several important stylized facts. Sharp time dependent lower and upper
bounds of global horizontal irradiations are estimated that improve the often
used physical bounds. The parameters of the beta distributed marginals of the
transformed data are allowed to be time dependent. A copula-based time series
model is introduced for the hourly and daily dependence structure based on a
simple graphical structure known from the theory of vine copulas. Non-Gaussian
copulas like Gumbel and BB1 copulas are used that allow for the important
feature of so-called tail dependence. Evaluation methods like the continuous
ranked probability score (CRPS), the energy score (ES) and the variogram score
(VS) are used to compare the power of the model for multivariate probabilistic
forecasting with other models used in the literature showing that our model
outperforms other models in many respects.
| stat |
On Efficient Design of Pilot Experiment for Generalized Linear Models | The experimental design for a generalized linear model (GLM) is important but
challenging since the design criterion often depends on model specification
including the link function, the linear predictor, and the unknown regression
coefficients. Prior to constructing locally or globally optimal designs, a
pilot experiment is usually conducted to provide some insights on the model
specifications. In pilot experiments, little information on the model
specification of GLM is available. Surprisingly, there is very limited research
on the design of pilot experiments for GLMs. In this work, we obtain some
theoretical understanding of the design efficiency in pilot experiments for
GLMs. Guided by the theory, we propose to adopt a low-discrepancy design with
respect to some target distribution for pilot experiments. The performance of
the proposed design is assessed through several numerical examples.
| stat |
Infinitesimal gradient boosting | We define infinitesimal gradient boosting as a limit of the popular
tree-based gradient boosting algorithm from machine learning. The limit is
considered in the vanishing-learning-rate asymptotic, that is when the learning
rate tends to zero and the number of gradient trees is rescaled accordingly.
For this purpose, we introduce a new class of randomized regression trees
bridging totally randomized trees and Extra Trees and using a softmax
distribution for binary splitting. Our main result is the convergence of the
associated stochastic algorithm and the characterization of the limiting
procedure as the unique solution of a nonlinear ordinary differential equation
in a infinite dimensional function space. Infinitesimal gradient boosting
defines a smooth path in the space of continuous functions along which the
training error decreases, the residuals remain centered and the total variation
is well controlled.
| stat |
Optimal quantization for nonuniform Cantor distributions | Let $P$ be a Borel probability measure on $\mathbb R$ such that $P=\frac 1 4
P\circ S_1^{-1} +\frac 3 4 P\circ S_2^{-1}$, where $S_1$ and $S_2$ are two
similarity mappings on $\mathbb R$ such that $S_1(x)=\frac 1 4 x $ and
$S_2(x)=\frac 1 2 x +\frac 12$ for all $x\in \mathbb R$. Such a probability
measure $P$ has support the Cantor set generated by $S_1$ and $S_2$. For this
probability measure, in this paper, we give an induction formula to determine
the optimal sets of $n$-means and the $n$th quantization errors for all $n\geq
2$. We have shown that the same induction formula also works for the Cantor
distribution $P:=\psi^2 P\circ S_1^{-1} +\psi^4 P\circ S_2^{-1}$ supported by
the Cantor set generated by $S_1(x)=\frac 13x$ and $S_2(x)=\frac 13 x+\frac 23$
for all $x\in \mathbb R$, where $\psi$ is the square root of the Golden ratio
$\frac 12(\sqrt 5-1)$. In addition, we give a counter example to show that the
induction formula does not work for all Cantor distributions. Using the
induction formula we obtain some results and observations which are also given
in this paper.
| stat |
Comparative analysis of the original and amplitude permutations | We compare the two basic ordinal patterns, i.e., the original and amplitude
permutations, used to characterize vector structures. The original permutation
consists of the indexes of reorganized values in the original vector. By
contrast, the amplitude permutation comprises the positions of values in the
reordered vector, and it directly reflects the temporal structure. To
accurately convey the structural characteristics of vectors, we modify indexes
of equal values in permutations to be the same as, for example, the smallest or
largest indexes in each group of equalities. Overall, we clarify the
relationship between the original and amplitude permutations. And the results
have implications for time- and amplitude-symmetric vectors and will lead to
further theoretical and experimental studies.
| stat |
Wasserstein GANs Work Because They Fail (to Approximate the Wasserstein
Distance) | Wasserstein GANs are based on the idea of minimising the Wasserstein distance
between a real and a generated distribution. We provide an in-depth
mathematical analysis of differences between the theoretical setup and the
reality of training Wasserstein GANs. In this work, we gather both theoretical
and empirical evidence that the WGAN loss is not a meaningful approximation of
the Wasserstein distance. Moreover, we argue that the Wasserstein distance is
not even a desirable loss function for deep generative models, and conclude
that the success of Wasserstein GANs can in truth be attributed to a failure to
approximate the Wasserstein distance.
| stat |
Forensic strength of evidence statements should preferably be likelihood
ratios calculated using relevant data, quantitative measurements, and
statistical models | Lennard (2013) [Fingerprint identification: how far have we come? Aus J
Forensic Sci. doi:10.1080/00450618.2012.752037] proposes that the numeric
output of statistical models should not be presented in court (except "if
necessary" / "if required"). Instead he argues in favour of an "expert opinion"
which may be informed by a statistical model but which is not itself the output
of a statistical model. We argue that his proposed procedure lacks the
transparency, the ease of testing of validity and reliability, and the relative
robustness to cognitive bias that are the strengths of a likelihood-ratio
approach based on relevant data, quantitative measurements, and statistical
models, and that the latter is therefore preferable.
| stat |
Matching Estimators for Causal Effects of Multiple Treatments | Matching estimators for average treatment effects are widely used in the
binary treatment setting, in which missing potential outcomes are imputed as
the average of observed outcomes of all matches for each unit. With more than
two treatment groups, however, estimation using matching requires additional
techniques. In this paper, we propose a nearest-neighbors matching estimator
for use with multiple, nominal treatments, and use simulations to show that
this method is precise and has coverage levels that are close to nominal. In
addition, we implement the proposed inference methods to examine the effects of
different medication regimens on long-term pain for patients experiencing motor
vehicle collision.
| stat |
Recovery of simultaneous low rank and two-way sparse coefficient
matrices, a nonconvex approach | We study the problem of recovery of matrices that are simultaneously low rank
and row and/or column sparse. Such matrices appear in recent applications in
cognitive neuroscience, imaging, computer vision, macroeconomics, and genetics.
We propose a GDT (Gradient Descent with hard Thresholding) algorithm to
efficiently recover matrices with such structure, by minimizing a bi-convex
function over a nonconvex set of constraints. We show linear convergence of the
iterates obtained by GDT to a region within statistical error of an optimal
solution. As an application of our method, we consider multi-task learning
problems and show that the statistical error rate obtained by GDT is near
optimal compared to minimax rate. Experiments demonstrate competitive
performance and much faster running speed compared to existing methods, on both
simulations and real data sets.
| stat |
Spatial CUSUM for Signal Region Detection | Detecting weak clustered signal in spatial data is important but challenging
in applications such as medical image and epidemiology. A more efficient
detection algorithm can provide more precise early warning, and effectively
reduce the decision risk and cost. To date, many methods have been developed to
detect signals with spatial structures. However, most of the existing methods
are either too conservative for weak signals or computationally too intensive.
In this paper, we consider a novel method named Spatial CUSUM (SCUSUM), which
employs the idea of the CUSUM procedure and false discovery rate controlling.
We develop theoretical properties of the method which indicates that
asymptotically SCUSUM can reach high classification accuracy. In the simulation
study, we demonstrate that SCUSUM is sensitive to weak spatial signals. This
new method is applied to a real fMRI dataset as illustration, and more
irregular weak spatial signals are detected in the images compared to some
existing methods, including the conventional FDR, FDR$_L$ and scan statistics.
| stat |
The Elliptical Processes: a Family of Fat-tailed Stochastic Processes | We present the elliptical processes -- a family of non-parametric
probabilistic models that subsumes the Gaussian process and the Student-t
process. This generalization includes a range of new fat-tailed behaviors yet
retains computational tractability. We base the elliptical processes on a
representation of elliptical distributions as a continuous mixture of Gaussian
distributions and derive closed-form expressions for the marginal and
conditional distributions. We perform numerical experiments on robust
regression using an elliptical process defined by a piecewise constant mixing
distribution, and show advantages compared with a Gaussian process. The
elliptical processes may become a replacement for Gaussian processes in several
settings, including when the likelihood is not Gaussian or when accurate tail
modeling is critical.
| stat |
False Discovery Rate for Functional Data | Since Benjamini and Hochberg introduced false discovery rate (FDR) in their
seminal paper, this has become a very popular approach to the multiple
comparisons problem. An increasingly popular topic within functional data
analysis is local inference, i.e., the continuous statistical testing of a null
hypothesis along the domain. The principal issue in this topic is the infinite
amount of tested hypotheses, which can be seen as an extreme case of the
multiple comparisons problem. In this paper we define and discuss the notion of
false discovery rate in a very general functional data setting. Moreover, a
continuous version of the Benjamini-Hochberg procedure is introduced along with
a definition of adjusted p-value function. Some general conditions are stated,
under which the functional Benjamini-Hochberg procedure provides control of the
functional FDR. Two different simulation studies are presented; the first study
has a one-dimensional domain and a comparison with another state of the art
method, and the second study has a planar two-dimensional domain. Finally, the
proposed method is applied to satellite measurements of Earth temperature. In
detail, we aim at identifying the regions of the planet where temperature has
significantly increased in the last decades. After adjustment, large areas are
still significant.
| stat |
Macroscopic Traffic Flow Modeling with Physics Regularized Gaussian
Process: A New Insight into Machine Learning Applications | Despite the wide implementation of machine learning (ML) techniques in
traffic flow modeling recently, those data-driven approaches often fall short
of accuracy in the cases with a small or noisy dataset. To address this issue,
this study presents a new modeling framework, named physics regularized machine
learning (PRML), to encode classical traffic flow models (referred as physical
models) into the ML architecture and to regularize the ML training process.
More specifically, a stochastic physics regularized Gaussian process (PRGP)
model is developed and a Bayesian inference algorithm is used to estimate the
mean and kernel of the PRGP. A physical regularizer based on macroscopic
traffic flow models is also developed to augment the estimation via a shadow GP
and an enhanced latent force model is used to encode physical knowledge into
stochastic processes. Based on the posterior regularization inference
framework, an efficient stochastic optimization algorithm is also developed to
maximize the evidence lowerbound of the system likelihood. To prove the
effectiveness of the proposed model, this paper conducts empirical studies on a
real-world dataset which is collected from a stretch of I-15 freeway, Utah.
Results show the new PRGP model can outperform the previous compatible methods,
such as calibrated pure physical models and pure machine learning methods, in
estimation precision and input robustness.
| stat |
Autoencoders Learn Generative Linear Models | We provide a series of results for unsupervised learning with autoencoders.
Specifically, we study shallow two-layer autoencoder architectures with shared
weights. We focus on three generative models for data that are common in
statistical machine learning: (i) the mixture-of-gaussians model, (ii) the
sparse coding model, and (iii) the sparsity model with non-negative
coefficients. For each of these models, we prove that under suitable choices of
hyperparameters, architectures, and initialization, autoencoders learned by
gradient descent can successfully recover the parameters of the corresponding
model. To our knowledge, this is the first result that rigorously studies the
dynamics of gradient descent for weight-sharing autoencoders. Our analysis can
be viewed as theoretical evidence that shallow autoencoder modules indeed can
be used as feature learning mechanisms for a variety of data models, and may
shed insight on how to train larger stacked architectures with autoencoders as
basic building blocks.
| stat |
Machine Learning's Dropout Training is Distributionally Robust Optimal | This paper shows that dropout training in Generalized Linear Models is the
minimax solution of a two-player, zero-sum game where an adversarial nature
corrupts a statistician's covariates using a multiplicative nonparametric
errors-in-variables model. In this game, nature's least favorable distribution
is dropout noise, where nature independently deletes entries of the covariate
vector with some fixed probability $\delta$. This result implies that dropout
training indeed provides out-of-sample expected loss guarantees for
distributions that arise from multiplicative perturbations of in-sample data.
In addition to the decision-theoretic analysis, the paper makes two more
contributions. First, there is a concrete recommendation on how to select the
tuning parameter $\delta$ to guarantee that, as the sample size grows large,
the in-sample loss after dropout training exceeds the true population loss with
some pre-specified probability. Second, the paper provides a novel,
parallelizable, Unbiased Multi-Level Monte Carlo algorithm to speed-up the
implementation of dropout training. Our algorithm has a much smaller
computational cost compared to the naive implementation of dropout, provided
the number of data points is much smaller than the dimension of the covariate
vector.
| stat |
On Arrhythmia Detection by Deep Learning and Multidimensional
Representation | An electrocardiogram (ECG) is a time-series signal that is represented by
one-dimensional (1-D) data. Higher dimensional representation contains more
information that is accessible for feature extraction. Hidden variables such as
frequency relation and morphology of segment is not directly accessible in the
time domain. In this paper, 1-D time series data is converted into
multi-dimensional representation in the form of multichannel 2-D images.
Following that, deep learning was used to train a deep neural network based
classifier to detect arrhythmias. The results of simulation on testing database
demonstrate the effectiveness of the proposed methodology by showing an
outstanding classification performance compared to other existing methods and
hand-crafted annotations made by certified cardiologists.
| stat |
Pycobra: A Python Toolbox for Ensemble Learning and Visualisation | We introduce \texttt{pycobra}, a Python library devoted to ensemble learning
(regression and classification) and visualisation. Its main assets are the
implementation of several ensemble learning algorithms, a flexible and generic
interface to compare and blend any existing machine learning algorithm
available in Python libraries (as long as a \texttt{predict} method is given),
and visualisation tools such as Voronoi tessellations. \texttt{pycobra} is
fully \texttt{scikit-learn} compatible and is released under the MIT
open-source license. \texttt{pycobra} can be downloaded from the Python Package
Index (PyPi) and Machine Learning Open Source Software (MLOSS). The current
version (along with Jupyter notebooks, extensive documentation, and continuous
integration tests) is available at
\href{https://github.com/bhargavvader/pycobra}{https://github.com/bhargavvader/pycobra}
and official documentation website is
\href{https://modal.lille.inria.fr/pycobra}{https://modal.lille.inria.fr/pycobra}.
| stat |
Risk & returns around FOMC press conferences: a novel perspective from
computer vision | I propose a new tool to characterize the resolution of uncertainty around
FOMC press conferences. It relies on the construction of a measure capturing
the level of discussion complexity between the Fed Chair and reporters during
the Q&A sessions. I show that complex discussions are associated with higher
equity returns and a drop in realized volatility. The method creates an
attention score by quantifying how much the Chair needs to rely on reading
internal documents to be able to answer a question. This is accomplished by
building a novel dataset of video images of the press conferences and
leveraging recent deep learning algorithms from computer vision. This
alternative data provides new information on nonverbal communication that
cannot be extracted from the widely analyzed FOMC transcripts. This paper can
be seen as a proof of concept that certain videos contain valuable information
for the study of financial markets.
| stat |
optimalFlow: Optimal-transport approach to flow cytometry gating and
population matching | Data obtained from Flow Cytometry present pronounced variability due to
biological and technical reasons. Biological variability is a well-known
phenomenon produced by measurements on different individuals, with different
characteristics such as illness, age, sex, etc. The use of different settings
for measurement, the variation of the conditions during experiments and the
different types of flow cytometers are some of the technical causes of
variability. This mixture of sources of variability makes the use of supervised
machine learning for identification of cell populations difficult. The present
work is conceived as a combination of strategies to facilitate the task of
supervised gating.
We propose $optimalFlowTemplates$, based on a similarity distance and
$\text{Wasserstein barycenters}$, which clusters cytometries and produces
prototype cytometries for the different groups. We show that supervised
learning, restricted to the new groups, performs better than the same
techniques applied to the whole collection. We also present
$optimalFlowClassification$, which uses a database of gated cytometries and
optimalFlowTemplates to assign cell types to a new cytometry. We show that this
procedure can outperform state of the art techniques in the proposed datasets.
Our code is freely available as $optimalFlow$ a Bioconductor R package at
https://bioconductor.org/packages/optimalFlow.
optimalFlowTemplates+optimalFlowClassification addresses the problem of using
supervised learning while accounting for biological and technical variability.
Our methodology provides a robust automated gating workflow that handles the
intrinsic variability of flow cytometry data well. Our main innovation is the
methodology itself and the optimal-transport techniques that we apply to flow
cytometry analysis.
| stat |
I-SPEC: An End-to-End Framework for Learning Transportable, Shift-Stable
Models | Shifts in environment between development and deployment cause classical
supervised learning to produce models that fail to generalize well to new
target distributions. Recently, many solutions which find invariant predictive
distributions have been developed. Among these, graph-based approaches do not
require data from the target environment and can capture more stable
information than alternative methods which find stable feature sets. However,
these approaches assume that the data generating process is known in the form
of a full causal graph, which is generally not the case. In this paper, we
propose I-SPEC, an end-to-end framework that addresses this shortcoming by
using data to learn a partial ancestral graph (PAG). Using the PAG we develop
an algorithm that determines an interventional distribution that is stable to
the declared shifts; this subsumes existing approaches which find stable
feature sets that are less accurate. We apply I-SPEC to a mortality prediction
problem to show it can learn a model that is robust to shifts without needing
upfront knowledge of the full causal DAG.
| stat |
Nonparametric Causal Feature Selection for Spatiotemporal Risk Mapping
of Malaria Incidence in Madagascar | Modern disease mapping draws upon a wealth of high resolution spatial data
products reflecting environmental and/or socioeconomic factors as covariates,
or `features', within a geostatistical framework to improve predictions of
disease risk. Feature selection is an important step in building these models,
helping to reduce overfitting and computational complexity, and to improve
model interpretability. Selecting only features that have a causal relationship
with the response variable could potentially improve predictions and
generalisability, but identifying these causal features from
non-interventional, spatiotemporal data is a challenging problem. Here we
examine the performance of a causal feature selection procedure with regard to
estimating malaria incidence in Madagascar. The studied procedure designed for
this task combines the PC algorithm with spatiotemporal prewhitening and
kernel-based independence tests extended to accommodate aggregated data. This
case study reveals a clear advantage for causal feature selection in terms of
the out-of-sample predictive accuracy in a forward temporal estimation task,
but not in a spatiotemporal interpolation task, in comparison with thresholded
spike-and-slab, for both linear and non-linear regression models. Compared to
no feature selection, causal feature selection was most beneficial in settings
wherein the volume of available data was low relative to the model complexity.
| stat |
Minimax Kernel Machine Learning for a Class of Doubly Robust Functionals | A moment function is called doubly robust if it is comprised of two nuisance
functions and the estimator based on it is a consistent estimator of the target
parameter even if one of the nuisance functions is misspecified. In this paper,
we consider a class of doubly robust moment functions originally introduced in
(Robins et al., 2008). We demonstrate that this moment function can be used to
construct estimating equations for the nuisance functions. The main idea is to
choose each nuisance function such that it minimizes the dependency of the
expected value of the moment function to the other nuisance function. We
implement this idea as a minimax optimization problem. We then provide
conditions required for asymptotic linearity of the estimator of the parameter
of interest, which are based on the convergence rate of the product of the
errors of the nuisance functions, as well as the local ill-posedness of a
conditional expectation operator. The convergence rates of the nuisance
functions are analyzed using the modern techniques in statistical learning
theory based on the Rademacher complexity of the function spaces. We
specifically focus on the case that the function spaces are reproducing kernel
Hilbert spaces, which enables us to use its spectral properties to analyze the
convergence rates. As an application of the proposed methodology, we consider
the parameter of average causal effect both in presence and absence of latent
confounders. For the case of presence of latent confounders, we use the
recently proposed proximal causal inference framework of (Miao et al., 2018;
Tchetgen Tchetgen et al., 2020), and hence our results lead to a robust
non-parametric estimator for average causal effect in this framework.
| stat |
Bayesian inference of Stochastic reaction networks using Multifidelity
Sequential Tempered Markov Chain Monte Carlo | Stochastic reaction network models are often used to explain and predict the
dynamics of gene regulation in single cells. These models usually involve
several parameters, such as the kinetic rates of chemical reactions, that are
not directly measurable and must be inferred from experimental data. Bayesian
inference provides a rigorous probabilistic framework for identifying these
parameters by finding a posterior parameter distribution that captures their
uncertainty. Traditional computational methods for solving inference problems
such as Markov Chain Monte Carlo methods based on classical Metropolis-Hastings
algorithm involve numerous serial evaluations of the likelihood function, which
in turn requires expensive forward solutions of the chemical master equation
(CME). We propose an alternative approach based on a multifidelity extension of
the Sequential Tempered Markov Chain Monte Carlo (ST-MCMC) sampler. This
algorithm is built upon Sequential Monte Carlo and solves the Bayesian
inference problem by decomposing it into a sequence of efficiently solved
subproblems that gradually increase model fidelity and the influence of the
observed data. We reformulate the finite state projection (FSP) algorithm, a
well-known method for solving the CME, to produce a hierarchy of surrogate
master equations to be used in this multifidelity scheme. To determine the
appropriate fidelity, we introduce a novel information-theoretic criteria that
seeks to extract the most information about the ultimate Bayesian posterior
from each model in the hierarchy without inducing significant bias. This novel
sampling scheme is tested with high performance computing resources using
biologically relevant problems.
| stat |
Streaming Variational Monte Carlo | Nonlinear state-space models are powerful tools to describe dynamical
structures in complex time series. In a streaming setting where data are
processed one sample at a time, simultaneous inference of the state and its
nonlinear dynamics has posed significant challenges in practice. We develop a
novel online learning framework, leveraging variational inference and
sequential Monte Carlo, which enables flexible and accurate Bayesian joint
filtering. Our method provides an approximation of the filtering posterior
which can be made arbitrarily close to the true filtering distribution for a
wide class of dynamics models and observation models. Specifically, the
proposed framework can efficiently approximate a posterior over the dynamics
using sparse Gaussian processes, allowing for an interpretable model of the
latent dynamics. Constant time complexity per sample makes our approach
amenable to online learning scenarios and suitable for real-time applications.
| stat |
What Do Neural Networks Learn When Trained With Random Labels? | We study deep neural networks (DNNs) trained on natural image data with
entirely random labels. Despite its popularity in the literature, where it is
often used to study memorization, generalization, and other phenomena, little
is known about what DNNs learn in this setting. In this paper, we show
analytically for convolutional and fully connected networks that an alignment
between the principal components of network parameters and data takes place
when training with random labels. We study this alignment effect by
investigating neural networks pre-trained on randomly labelled image data and
subsequently fine-tuned on disjoint datasets with random or real labels. We
show how this alignment produces a positive transfer: networks pre-trained with
random labels train faster downstream compared to training from scratch even
after accounting for simple effects, such as weight scaling. We analyze how
competing effects, such as specialization at later layers, may hide the
positive transfer. These effects are studied in several network architectures,
including VGG16 and ResNet18, on CIFAR10 and ImageNet.
| stat |
Differentiable Expected Hypervolume Improvement for Parallel
Multi-Objective Bayesian Optimization | In many real-world scenarios, decision makers seek to efficiently optimize
multiple competing objectives in a sample-efficient fashion. Multi-objective
Bayesian optimization (BO) is a common approach, but many of the
best-performing acquisition functions do not have known analytic gradients and
suffer from high computational overhead. We leverage recent advances in
programming models and hardware acceleration for multi-objective BO using
Expected Hypervolume Improvement (EHVI)---an algorithm notorious for its high
computational complexity. We derive a novel formulation of q-Expected
Hypervolume Improvement (qEHVI), an acquisition function that extends EHVI to
the parallel, constrained evaluation setting. qEHVI is an exact computation of
the joint EHVI of q new candidate points (up to Monte-Carlo (MC) integration
error). Whereas previous EHVI formulations rely on gradient-free acquisition
optimization or approximated gradients, we compute exact gradients of the MC
estimator via auto-differentiation, thereby enabling efficient and effective
optimization using first-order and quasi-second-order methods. Our empirical
evaluation demonstrates that qEHVI is computationally tractable in many
practical scenarios and outperforms state-of-the-art multi-objective BO
algorithms at a fraction of their wall time.
| stat |
Deep Learning Estimation of Absorbed Dose for Nuclear Medicine
Diagnostics | The distribution of energy dose from Lu$^{177}$ radiotherapy can be estimated
by convolving an image of a time-integrated activity distribution with a dose
voxel kernel (DVK) consisting of different types of tissues. This fast and
inacurate approximation is inappropriate for personalized dosimetry as it
neglects tissue heterogenity. The latter can be calculated using different
imaging techniques such as CT and SPECT combined with a time consuming
monte-carlo simulation. The aim of this study is, for the first time, an
estimation of DVKs from CT-derived density kernels (DK) via deep learning in
convolutional neural networks (CNNs). The proposed CNN achieved, on the test
set, a mean intersection over union (IOU) of $= 0.86$ after $308$ epochs and a
corresponding mean squared error (MSE) $= 1.24 \cdot 10^{-4}$. This
generalization ability shows that the trained CNN can indeed learn the
difficult transfer function from DK to DVK. Future work will evaluate DVKs
estimated by CNNs with full monte-carlo simulations of a whole body CT to
predict patient specific voxel dose maps.
| stat |
Causal Mediation Analysis for Sparse and Irregular Longitudinal Data | Causal mediation analysis seeks to investigate how the treatment effect of an
exposure on outcomes is mediated through intermediate variables. Although many
applications involve longitudinal data, the existing methods are not directly
applicable to settings where the mediator and outcome are measured on sparse
and irregular time grids. We extend the existing causal mediation framework
from a functional data analysis perspective, viewing the sparse and irregular
longitudinal data as realizations of underlying smooth stochastic processes. We
define causal estimands of direct and indirect effects accordingly and provide
corresponding identification assumptions. For estimation and inference, we
employ a functional principal component analysis approach for dimension
reduction and use the first few functional principal components instead of the
whole trajectories in the structural equation models. We adopt the Bayesian
paradigm to accurately quantify the uncertainties. The operating
characteristics of the proposed methods are examined via simulations. We apply
the proposed methods to a longitudinal data set from a wild baboon population
in Kenya to investigate the causal relationships between early adversity,
strength of social bonds between animals, and adult glucocorticoid hormone
concentrations. We find that early adversity has a significant direct effect (a
9-14% increase) on females' glucocorticoid concentrations across adulthood, but
find little evidence that these effects were mediated by weak social bonds.
| stat |
A Debiased MDI Feature Importance Measure for Random Forests | Tree ensembles such as Random Forests have achieved impressive empirical
success across a wide variety of applications. To understand how these models
make predictions, people routinely turn to feature importance measures
calculated from tree ensembles. It has long been known that Mean Decrease
Impurity (MDI), one of the most widely used measures of feature importance,
incorrectly assigns high importance to noisy features, leading to systematic
bias in feature selection. In this paper, we address the feature selection bias
of MDI from both theoretical and methodological perspectives. Based on the
original definition of MDI by Breiman et al. for a single tree, we derive a
tight non-asymptotic bound on the expected bias of MDI importance of noisy
features, showing that deep trees have higher (expected) feature selection bias
than shallow ones. However, it is not clear how to reduce the bias of MDI using
its existing analytical expression. We derive a new analytical expression for
MDI, and based on this new expression, we are able to propose a debiased MDI
feature importance measure using out-of-bag samples, called MDI-oob. For both
the simulated data and a genomic ChIP dataset, MDI-oob achieves
state-of-the-art performance in feature selection from Random Forests for both
deep and shallow trees.
| stat |
Properties of the Stochastic Approximation EM Algorithm with Mini-batch
Sampling | To deal with very large datasets a mini-batch version of the Monte Carlo
Markov Chain Stochastic Approximation Expectation-Maximization algorithm for
general latent variable models is proposed. For exponential models the
algorithm is shown to be convergent under classicalconditions as the number of
iterations increases. Numerical experiments illustrate the performance of the
mini-batch algorithm in various models.In particular, we highlight that
mini-batch sampling results in an important speed-up of the convergence of the
sequence of estimators generated by the algorithm. Moreover, insights on the
effect of the mini-batch size on the limit distribution are presented. Finally,
we illustrate how to use mini-batch sampling in practice to improve results
when a constraint on the computing time is given.
| stat |
Expressive Graph Informer Networks | Applying machine learning to molecules is challenging because of their
natural representation as graphs rather than vectors.Several architectures have
been recently proposed for deep learning from molecular graphs, but they suffer
from informationbottlenecks because they only pass information from a graph
node to its direct neighbors. Here, we introduce a more expressiveroute-based
multi-attention mechanism that incorporates features from routes between node
pairs. We call the resulting methodGraph Informer. A single network layer can
therefore attend to nodes several steps away. We show empirically that the
proposedmethod compares favorably against existing approaches in two prediction
tasks: (1) 13C Nuclear Magnetic Resonance (NMR)spectra, improving the
state-of-the-art with an MAE of 1.35 ppm and (2) predicting drug bioactivity
and toxicity. Additionally, wedevelop a variant called injective Graph Informer
that isprovablyas powerful as the Weisfeiler-Lehman test for graph
isomorphism.Furthermore, we demonstrate that the route information allows the
method to be informed about thenonlocal topologyof the graphand, thus, even go
beyond the capabilities of the Weisfeiler-Lehman test.
| stat |
Learning with Bounded Instance- and Label-dependent Label Noise | Instance- and Label-dependent label Noise (ILN) widely exists in real-world
datasets but has been rarely studied. In this paper, we focus on Bounded
Instance- and Label-dependent label Noise (BILN), a particular case of ILN
where the label noise rates -- the probabilities that the true labels of
examples flip into the corrupted ones -- have upper bound less than $1$.
Specifically, we introduce the concept of distilled examples, i.e. examples
whose labels are identical with the labels assigned for them by the Bayes
optimal classifier, and prove that under certain conditions classifiers learnt
on distilled examples will converge to the Bayes optimal classifier. Inspired
by the idea of learning with distilled examples, we then propose a learning
algorithm with theoretical guarantees for its robustness to BILN. At last,
empirical evaluations on both synthetic and real-world datasets show
effectiveness of our algorithm in learning with BILN.
| stat |
Estimating Basis Functions in Massive Fields under the Spatial Mixed
Effects Model | Spatial prediction is commonly achieved under the assumption of a Gaussian
random field (GRF) by obtaining maximum likelihood estimates of parameters, and
then using the kriging equations to arrive at predicted values. For massive
datasets, fixed rank kriging using the Expectation-Maximization (EM) algorithm
for estimation has been proposed as an alternative to the usual but
computationally prohibitive kriging method. The method reduces computation cost
of estimation by redefining the spatial process as a linear combination of
basis functions and spatial random effects. A disadvantage of this method is
that it imposes constraints on the relationship between the observed locations
and the knots. We develop an alternative method that utilizes the Spatial Mixed
Effects (SME) model, but allows for additional flexibility by estimating the
range of the spatial dependence between the observations and the knots via an
Alternating Expectation Conditional Maximization (AECM) algorithm. Experiments
show that our methodology improves estimation without sacrificing prediction
accuracy while also minimizing the additional computational burden of extra
parameter estimation. The methodology is applied to a temperature data set
archived by the United States National Climate Data Center, with improved
results over previous methodology.
| stat |
Statistics and Samples in Distributional Reinforcement Learning | We present a unifying framework for designing and analysing distributional
reinforcement learning (DRL) algorithms in terms of recursively estimating
statistics of the return distribution. Our key insight is that DRL algorithms
can be decomposed as the combination of some statistical estimator and a method
for imputing a return distribution consistent with that set of statistics. With
this new understanding, we are able to provide improved analyses of existing
DRL algorithms as well as construct a new algorithm (EDRL) based upon
estimation of the expectiles of the return distribution. We compare EDRL with
existing methods on a variety of MDPs to illustrate concrete aspects of our
analysis, and develop a deep RL variant of the algorithm, ER-DQN, which we
evaluate on the Atari-57 suite of games.
| stat |
Detecting linear trend changes and point anomalies in data sequences | We propose TrendSegment, a methodology for detecting multiple change-points
corresponding to linear trend changes or point anomalies in one dimensional
data. A core ingredient of TrendSegment is a new Tail-Greedy Unbalanced Wavelet
transform: a conditionally orthonormal, bottom-up transformation of the data
through an adaptively constructed unbalanced wavelet basis, which results in a
sparse representation of the data. The bottom-up nature of this multiscale
decomposition enables the detection of point anomalies and linear trend changes
at once as the decomposition focuses on local features in its early stages and
on global features next. To reduce the computational complexity, the proposed
method merges multiple regions in a single pass over the data. We show the
consistency of the estimated number and locations of change-points. The
practicality of our approach is demonstrated through simulations and two real
data examples, involving Iceland temperature data and sea ice extent of the
Arctic and the Antarctic. Our methodology is implemented in the R package
trendsegmentR, available from CRAN.
| stat |
Fast Multivariate Log-Concave Density Estimation | A novel computational approach to log-concave density estimation is proposed.
Previous approaches utilize the piecewise-affine parametrization of the density
induced by the given sample set. The number of parameters as well as non-smooth
subgradient-based convex optimization for determining the maximum likelihood
density estimate cause long runtimes for dimensions $d \geq 2$ and large sample
sets. The presented approach is based on mildly non-convex smooth
approximations of the objective function and \textit{sparse}, adaptive
piecewise-affine density parametrization. Established memory-efficient
numerical optimization techniques enable to process larger data sets for
dimensions $d \geq 2$. While there is no guarantee that the algorithm returns
the maximum likelihood estimate for every problem instance, we provide
comprehensive numerical evidence that it does yield near-optimal results after
significantly shorter runtimes. For example, 10000 samples in $\mathbb{R}^2$
are processed in two seconds, rather than in $\approx 14$ hours required by the
previous approach to terminate. For higher dimensions, density estimation
becomes tractable as well: Processing $10000$ samples in $\mathbb{R}^6$
requires 35 minutes. The software is publicly available as CRAN R package
fmlogcondens.
| stat |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.