title
stringlengths 6
244
| abstract
stringlengths 19
5.09k
| label
stringclasses 10
values |
|---|---|---|
A simple consistent Bayes factor for testing the Kendall rank
correlation coefficient | In this paper, we propose a simple and easy-to-implement Bayesian hypothesis
test for the presence of an association, described by Kendall's tau
coefficient, between two variables measured on at least an ordinal scale. Owing
to the absence of the likelihood functions for the data, we employ the
asymptotic sampling distributions of the test statistic as the working
likelihoods and then specify a truncated normal prior distribution on the
noncentrality parameter of the alternative hypothesis, which results in the
Bayes factor available in closed form in terms of the cumulative distribution
function of the standard normal distribution. Investigating the asymptotic
behavior of the Bayes factor we find the conditions of the priors so that it is
consistent to whichever the hypothesis is true. Simulation studies and a
real-data application are used to illustrate the effectiveness of the proposed
Bayes factor. It deserves mentioning that the proposed method can be easily
covered in undergraduate and graduate courses in nonparametric statistics with
an emphasis on students' Bayesian thinking for data analysis.
| stat |
A Stein Goodness-of-fit Test for Directional Distributions | In many fields, data appears in the form of direction (unit vector) and usual
statistical procedures are not applicable to such directional data. In this
study, we propose non-parametric goodness-of-fit testing procedures for general
directional distributions based on kernel Stein discrepancy. Our method is
based on Stein's operator on spheres, which is derived by using Stokes'
theorem. Notably, the proposed method is applicable to distributions with an
intractable normalization constant, which commonly appear in directional
statistics. Experimental results demonstrate that the proposed methods control
type-I error well and have larger power than existing tests, including the test
based on the maximum mean discrepancy.
| stat |
Spherical Principal Curves | This paper presents a new approach for dimension reduction of data observed
in a sphere. Several dimension reduction techniques have recently developed for
the analysis of non-Euclidean data. As a pioneer work, Hauberg (2016) attempted
to implement principal curves on Riemannian manifolds. However, this approach
uses approximations to deal with data on Riemannian manifolds, which causes
distorted results. In this study, we propose a new approach to construct
principal curves on a sphere by a projection of the data onto a continuous
curve. Our approach lies in the same line of Hastie and Stuetzle (1989) that
proposed principal curves for Euclidean space data. We further investigate the
stationarity of the proposed principal curves that satisfy the self-consistency
on a sphere. Results from real data analysis with earthquake data and
simulation examples demonstrate the promising empirical properties of the
proposed approach.
| stat |
Forest Guided Smoothing | We use the output of a random forest to define a family of local smoothers
with spatially adaptive bandwidth matrices. The smoother inherits the
flexibility of the original forest but, since it is a simple, linear smoother,
it is very interpretable and it can be used for tasks that would be intractable
for the original forest. This includes bias correction, confidence intervals,
assessing variable importance and methods for exploring the structure of the
forest. We illustrate the method on some synthetic examples and on data related
to Covid-19.
| stat |
Dirichlet Simplex Nest and Geometric Inference | We propose Dirichlet Simplex Nest, a class of probabilistic models suitable
for a variety of data types, and develop fast and provably accurate inference
algorithms by accounting for the model's convex geometry and low dimensional
simplicial structure. By exploiting the connection to Voronoi tessellation and
properties of Dirichlet distribution, the proposed inference algorithm is shown
to achieve consistency and strong error bound guarantees on a range of model
settings and data distributions. The effectiveness of our model and the
learning algorithm is demonstrated by simulations and by analyses of text and
financial data.
| stat |
A proof of convergence of multi-class logistic regression network | This paper revisits the special type of a neural network known under two
names. In the statistics and machine learning community it is known as a
multi-class logistic regression neural network. In the neural network
community, it is simply the soft-max layer. The importance is underscored by
its role in deep learning: as the last layer, whose autput is actually the
classification of the input patterns, such as images. Our exposition focuses on
mathematically rigorous derivation of the key equation expressing the gradient.
The fringe benefit of our approach is a fully vectorized expression, which is a
basis of an efficient implementation. The second result of this paper is the
positivity of the second derivative of the cross-entropy loss function as
function of the weights. This result proves that optimization methods based on
convexity may be used to train this network. As a corollary, we demonstrate
that no $L^2$-regularizer is needed to guarantee convergence of gradient
descent.
| stat |
The Duration of Optimal Stopping Problems | Optimal stopping problems give rise to random distributions describing how
many applicants the decision-maker will sample or interview before choosing
one, a quantity sometimes referred to as the search time or process duration.
This research note surveys several variants of optimal stopping problems,
extends earlier results in various directions, and shows how many interviews
are expected to be conducted in various settings. The focus is on problems that
require a decision-maker to choose a candidate from a pool of sequential
applicants with no recall, in the vein of previously studied Cayley-Moser,
Secretary and Sultan's Dowry problems.
| stat |
Selective prediction-set models with coverage guarantees | Though black-box predictors are state-of-the-art for many complex tasks, they
often fail to properly quantify predictive uncertainty and may provide
inappropriate predictions for unfamiliar data. Instead, we can learn more
reliable models by letting them either output a prediction set or abstain when
the uncertainty is high. We propose training these selective prediction-set
models using an uncertainty-aware loss minimization framework, which unifies
ideas from decision theory and robust maximum likelihood. Moreover, since
black-box methods are not guaranteed to output well-calibrated prediction sets,
we show how to calculate point estimates and confidence intervals for the true
coverage of any selective prediction-set model, as well as a uniform mixture of
K set models obtained from K-fold sample-splitting. When applied to predicting
in-hospital mortality and length-of-stay for ICU patients, our model
outperforms existing approaches on both in-sample and out-of-sample age groups,
and our recalibration method provides accurate inference for prediction set
coverage.
| stat |
Projected Statistical Methods for Distributional Data on the Real Line
with the Wasserstein Metric | We present a novel class of projected methods, to perform statistical
analysis on a data set of probability distributions on the real line, with the
2-Wasserstein metric. We focus in particular on Principal Component Analysis
(PCA) and regression. To define these models, we exploit a representation of
the Wasserstein space closely related to its weak Riemannian structure, by
mapping the data to a suitable linear space and using a metric projection
operator to constrain the results in the Wasserstein space. By carefully
choosing the tangent point, we are able to derive fast empirical methods,
exploiting a constrained B-spline approximation. As a byproduct of our
approach, we are also able to derive faster routines for previous work on PCA
for distributions. By means of simulation studies, we compare our approaches to
previously proposed methods, showing that our projected PCA has similar
performance for a fraction of the computational cost and that the projected
regression is extremely flexible even under misspecification. Several
theoretical properties of the models are investigated and asymptotic
consistency is proven. Two real world applications to Covid-19 mortality in the
US and wind speed forecasting are discussed.
| stat |
Tree boosting for learning probability measures | Learning probability measures based on an i.i.d. sample is a fundamental
inference task, but is challenging when the sample space is high-dimensional.
Inspired by the success of tree boosting in high-dimensional classification and
regression, we propose a tree boosting method for learning high-dimensional
probability distributions. We formulate concepts of "addition" and "residuals"
on probability distributions in terms of compositions of a new, more general
notion of multivariate cumulative distribution functions (CDFs) than classical
CDFs. This then gives rise to a simple boosting algorithm based on
forward-stagewise (FS) fitting of an additive ensemble of measures, which
sequentially minimizes the entropy loss. The output of the FS algorithm allows
analytic computation of the probability density function for the fitted
distribution. It also provides an exact simulator for drawing independent Monte
Carlo samples from the fitted measure. Typical considerations in applying
boosting--namely choosing the number of trees, setting the appropriate level of
shrinkage/regularization in the weak learner, and the evaluation of variable
importance--can all be accomplished in an analogous fashion to traditional
boosting in supervised learning. Numerical experiments confirm that boosting
can substantially improve the fit to multivariate distributions compared to the
state-of-the-art single-tree learner and is computationally efficient. We
illustrate through an application to a data set from mass cytometry how the
simulator can be used to investigate various aspects of the underlying
distribution.
| stat |
Modelling Grocery Retail Topic Distributions: Evaluation,
Interpretability and Stability | Understanding the shopping motivations behind market baskets has high
commercial value in the grocery retail industry. Analyzing shopping
transactions demands techniques that can cope with the volume and
dimensionality of grocery transactional data while keeping interpretable
outcomes. Latent Dirichlet Allocation (LDA) provides a suitable framework to
process grocery transactions and to discover a broad representation of
customers' shopping motivations. However, summarizing the posterior
distribution of an LDA model is challenging, while individual LDA draws may not
be coherent and cannot capture topic uncertainty. Moreover, the evaluation of
LDA models is dominated by model-fit measures which may not adequately capture
the qualitative aspects such as interpretability and stability of topics.
In this paper, we introduce clustering methodology that post-processes
posterior LDA draws to summarise the entire posterior distribution and identify
semantic modes represented as recurrent topics. Our approach is an alternative
to standard label-switching techniques and provides a single posterior summary
set of topics, as well as associated measures of uncertainty. Furthermore, we
establish a more holistic definition for model evaluation, which assesses topic
models based not only on their likelihood but also on their coherence,
distinctiveness and stability. By means of a survey, we set thresholds for the
interpretation of topic coherence and topic similarity in the domain of grocery
retail data. We demonstrate that the selection of recurrent topics through our
clustering methodology not only improves model likelihood but also outperforms
the qualitative aspects of LDA such as interpretability and stability. We
illustrate our methods on an example from a large UK supermarket chain.
| stat |
Statistical models for short and long term forecasts of snow depth | Forecasting of future snow depths is useful for many applications like road
safety, winter sport activities, avalanche risk assessment and hydrology.
Motivated by the lack of statistical forecasts models for snow depth, in this
paper we present a set of models to fill this gap. First, we present a model to
do short term forecasts when we assume that reliable weather forecasts of air
temperature and precipitation are available. The covariates are included
nonlinearly into the model following basic physical principles of snowfall,
snow aging and melting. Due to the large set of observations with snow depth
equal to zero, we use a zero-inflated gamma regression model, which is commonly
used to similar applications like precipitation. We also do long term forecasts
of snow depth and much further than traditional weather forecasts for
temperature and precipitation. The long-term forecasts are based on fitting
models to historic time series of precipitation, temperature and snow depth. We
fit the models to data from three locations in Norway with different climatic
properties. Forecasting five days into the future, the results showed that,
given reliable weather forecasts of temperature and precipitation, the forecast
errors in absolute value was between 3 and 7 cm for different locations in
Norway. Forecasting three weeks into the future, the forecast errors were
between 7 and 16 cm.
| stat |
A simple method for implementing Monte Carlo tests | We consider a statistical test whose p-value can only be approximated using
Monte Carlo simulations. We are interested in deciding whether the p-value for
an observed data set lies above or below a given threshold such as 5%. We want
to ensure that the resampling risk, the probability of the (Monte Carlo)
decision being different from the true decision, is uniformly bounded. This
article introduces a simple open-ended method with this property, the
confidence sequence method (CSM). We compare our approach to another algorithm,
SIMCTEST, which also guarantees an (asymptotic) uniform bound on the resampling
risk, as well as to other Monte Carlo procedures without a uniform bound. CSM
is free of tuning parameters and conservative. It has the same theoretical
guarantee as SIMCTEST and, in many settings, similar stopping boundaries. As it
is much simpler than other methods, CSM is a useful method for practical
applications.
| stat |
A Tutorial on Multivariate $k$-Statistics and their Computation | This document aims to provide an accessible tutorial on the unbiased
estimation of multivariate cumulants, using $k$-statistics. We offer an
explicit and general formula for multivariate $k$-statistics of arbitrary
order. We also prove that the $k$-statistics are unbiased, using M\"obius
inversion and rudimentary combinatorics. Many detailed examples are considered
throughout the paper. We conclude with a discussion of $k$-statistics
computation, including the challenge of time complexity, and we examine a
couple of possible avenues to improve the efficiency of this computation. The
purpose of this document is threefold: to provide a clear introduction to
$k$-statistics without relying on specialized tools like the umbral calculus;
to construct an explicit formula for $k$-statistics that might facilitate
future approximations and faster algorithms; and to serve as a companion paper
to our Python library PyMoments, which implements this formula.
| stat |
Probabilistic Recalibration of Forecasts | We present a scheme by which a probabilistic forecasting system whose
predictions have poor probabilistic calibration may be recalibrated by
incorporating past performance information to produce a new forecasting system
that is demonstrably superior to the original, in that one may use it to
consistently win wagers against someone using the original system. The scheme
utilizes Gaussian process (GP) modeling to estimate a probability distribution
over the Probability Integral Transform (PIT) of a scalar predictand. The GP
density estimate gives closed-form access to information entropy measures
associated with the estimated distribution, which allows prediction of winnings
in wagers against the base forecasting system. A separate consequence of the
procedure is that the recalibrated forecast has a uniform expected PIT
distribution. A distinguishing feature of the procedure is that it is
appropriate even if the PIT values are not i.i.d. The recalibration scheme is
formulated in a framework that exploits the deep connections between
information theory, forecasting, and betting. We demonstrate the effectiveness
of the scheme in two case studies: a laboratory experiment with a nonlinear
circuit and seasonal forecasts of the intensity of the El Ni\~no-Southern
Oscillation phenomenon.
| stat |
Kernel-Distance-Based Covariate Balancing | A common concern in observational studies focuses on properly evaluating the
causal effect, which usually refers to the average treatment effect or the
average treatment effect on the treated. In this paper, we propose a data
preprocessing method, the Kernel-distance-based covariate balancing, for
observational studies with binary treatments. This proposed method yields a set
of unit weights for the treatment and control groups, respectively, such that
the reweighted covariate distributions can satisfy a set of pre-specified
balance conditions. This preprocessing methodology can effectively reduce
confounding bias of subsequent estimation of causal effects. We demonstrate the
implementation and performance of Kernel-distance-based covariate balancing
with Monte Carlo simulation experiments and a real data analysis.
| stat |
Social Network Mediation Analysis: a Latent Space Approach | Social networks contain data on both actor attributes and social connections
among them. Such connections reflect the dependence among social actors, which
is important for individual's mental health and social development. To
investigate the potential mediation role of a social network, we propose a
mediation model with a social network as a mediator. In the model, dependence
among actors is accounted by a few mutually orthogonal latent dimensions. The
scores on these dimensions are directly involved in the intervention process
between an independent variable and a dependent variable. Because all the
latent dimensions are equivalent in terms of their relationship to social
networks, it is hardly to name them. The intervening effect through an
individual dimension is thus of little practical interest. Therefore, we would
rather focus on the mediation effect of a network. Although the scores are not
unique, we rigorously articulate that the proposed network mediation effect is
still well-defined. To estimate the model, we adopt a Bayesian estimation
method. This modeling framework and the Bayesian estimation method is evaluated
through a simulation study under representative conditions. Its usefulness is
demonstrated through an empirical application to a college friendship network.
| stat |
Precision-Recall Curve (PRC) Classification Trees | The classification of imbalanced data has presented a significant challenge
for most well-known classification algorithms that were often designed for data
with relatively balanced class distributions. Nevertheless skewed class
distribution is a common feature in real world problems. It is especially
prevalent in certain application domains with great need for machine learning
and better predictive analysis such as disease diagnosis, fraud detection,
bankruptcy prediction, and suspect identification. In this paper, we propose a
novel tree-based algorithm based on the area under the precision-recall curve
(AUPRC) for variable selection in the classification context. Our algorithm,
named as the "Precision-Recall Curve classification tree", or simply the "PRC
classification tree" modifies two crucial stages in tree building. The first
stage is to maximize the area under the precision-recall curve in node variable
selection. The second stage is to maximize the harmonic mean of recall and
precision (F-measure) for threshold selection. We found the proposed PRC
classification tree, and its subsequent extension, the PRC random forest, work
well especially for class-imbalanced data sets. We have demonstrated that our
methods outperform their classic counterparts, the usual CART and random forest
for both synthetic and real data. Furthermore, the ROC classification tree
proposed by our group previously has shown good performance in imbalanced data.
The combination of them, the PRC-ROC tree, also shows great promise in
identifying the minority class.
| stat |
Strong Consistency of Spectral Clustering for Stochastic Block Models | In this paper we prove the strong consistency of several methods based on the
spectral clustering techniques that are widely used to study the community
detection problem in stochastic block models (SBMs). We show that under some
weak conditions on the minimal degree, the number of communities, and the
eigenvalues of the probability block matrix, the K-means algorithm applied to
the eigenvectors of the graph Laplacian associated with its first few largest
eigenvalues can classify all individuals into the true community uniformly
correctly almost surely. Extensions to both regularized spectral clustering and
degree-corrected SBMs are also considered. We illustrate the performance of
different methods on simulated networks.
| stat |
Computational Causal Inference | We introduce computational causal inference as an interdisciplinary field
across causal inference, algorithms design and numerical computing. The field
aims to develop software specializing in causal inference that can analyze
massive datasets with a variety of causal effects, in a performant, general,
and robust way. The focus on software improves research agility, and enables
causal inference to be easily integrated into large engineering systems. In
particular, we use computational causal inference to deepen the relationship
between causal inference, online experimentation, and algorithmic decision
making.
This paper describes the new field, the demand, opportunities for
scalability, open challenges, and begins the discussion for how the community
can unite to solve challenges for scaling causal inference and decision making.
| stat |
Comparing high dimensional partitions, with the Coclustering Adjusted
Rand Index | We consider the simultaneous clustering of rows and columns of a matrix and
more particularly the ability to measure the agreement between two
co-clustering partitions. The new criterion we developed is based on the
Adjusted Rand Index and is called the Co-clustering Adjusted Rand Index named
CARI. We also suggest new improvements to existing criteria such as the
Classification Error which counts the proportion of misclassified cells and the
Extended Normalized Mutual Information criterion which is a generalization of
the criterion based on mutual information in the case of classic
classifications. We study these criteria with regard to some desired properties
deriving from the co-clustering context. Experiments on simulated and real
observed data are proposed to compare the behavior of these criteria.
| stat |
Convergence of Parameter Estimates for Regularized Mixed Linear
Regression Models | We consider {\em Mixed Linear Regression (MLR)}, where training data have
been generated from a mixture of distinct linear models (or clusters) and we
seek to identify the corresponding coefficient vectors. We introduce a {\em
Mixed Integer Programming (MIP)} formulation for MLR subject to regularization
constraints on the coefficient vectors. We establish that as the number of
training samples grows large, the MIP solution converges to the true
coefficient vectors in the absence of noise. Subject to slightly stronger
assumptions, we also establish that the MIP identifies the clusters from which
the training samples were generated. In the special case where training data
come from a single cluster, we establish that the corresponding MIP yields a
solution that converges to the true coefficient vector even when training data
are perturbed by (martingale difference) noise. We provide a counterexample
indicating that in the presence of noise, the MIP may fail to produce the true
coefficient vectors for more than one clusters. We also provide numerical
results testing the MIP solutions in synthetic examples with noise.
| stat |
Sparse inversion for derivative of log determinant | Algorithms for Gaussian process, marginal likelihood methods or restricted
maximum likelihood methods often require derivatives of log determinant terms.
These log determinants are usually parametric with variance parameters of the
underlying statistical models. This paper demonstrates that, when the
underlying matrix is sparse, how to take the advantage of sparse
inversion---selected inversion which share the same sparsity as the original
matrix---to accelerate evaluating the derivative of log determinant.
| stat |
Learning the Stein Discrepancy for Training and Evaluating Energy-Based
Models without Sampling | We present a new method for evaluating and training unnormalized density
models. Our approach only requires access to the gradient of the unnormalized
model's log-density. We estimate the Stein discrepancy between the data density
$p(x)$ and the model density $q(x)$ defined by a vector function of the data.
We parameterize this function with a neural network and fit its parameters to
maximize the discrepancy. This yields a novel goodness-of-fit test which
outperforms existing methods on high dimensional data. Furthermore, optimizing
$q(x)$ to minimize this discrepancy produces a novel method for training
unnormalized models which scales more gracefully than existing methods. The
ability to both learn and compare models is a unique feature of the proposed
method.
| stat |
Metric-Optimized Example Weights | Real-world machine learning applications often have complex test metrics, and
may have training and test data that are not identically distributed. Motivated
by known connections between complex test metrics and cost-weighted learning,
we propose addressing these issues by using a weighted loss function with a
standard loss, where the weights on the training examples are learned to
optimize the test metric on a validation set. These metric-optimized example
weights can be learned for any test metric, including black box and customized
ones for specific applications. We illustrate the performance of the proposed
method on diverse public benchmark datasets and real-world applications. We
also provide a generalization bound for the method.
| stat |
Outer power transformations of hierarchical Archimedean copulas:
Construction, sampling and estimation | A large number of commonly used parametric Archimedean copula (AC) families
are restricted to a single parameter, connected to a concordance measure such
as Kendall's tau. This often leads to poor statistical fits, particularly in
the joint tails, and can sometimes even limit the ability to model concordance
or tail dependence mathematically. This work suggests outer power (OP)
transformations of Archimedean generators to overcome these limitations. The
copulas generated by OP-transformed generators can, for example, allow one to
capture both a given concordance measure and a tail dependence coefficient
simultaneously. For exchangeable OP-transformed ACs, a formula for computing
tail dependence coefficients is obtained, as well as two feasible OP AC
estimators are proposed and their properties studied by simulation. For
hierarchical extensions of OP-transformed ACs, a new construction principle,
efficient sampling and parameter estimation are addressed. By simulation,
convergence rate and standard errors of the proposed estimator are studied.
Excellent tail fitting capabilities of OP-transformed hierarchical AC models
are demonstrated in a risk management application. The results show that the OP
transformation is able to improve the statistical fit of exchangeable ACs,
particularly of those that cannot capture upper tail dependence or strong
concordance, as well as the statistical fit of hierarchical ACs, especially in
terms of tail dependence and higher dimensions. Given how comparably simple it
is to include OP transformations into existing exchangeable and hierarchical AC
models, this transformation provides an attractive trade-off between
computational effort and statistical improvement.
| stat |
Trading Convergence Rate with Computational Budget in High Dimensional
Bayesian Optimization | Scaling Bayesian optimisation (BO) to high-dimensional search spaces is a
active and open research problems particularly when no assumptions are made on
function structure. The main reason is that at each iteration, BO requires to
find global maximisation of acquisition function, which itself is a non-convex
optimization problem in the original search space. With growing dimensions, the
computational budget for this maximisation gets increasingly short leading to
inaccurate solution of the maximisation. This inaccuracy adversely affects both
the convergence and the efficiency of BO. We propose a novel approach where the
acquisition function only requires maximisation on a discrete set of low
dimensional subspaces embedded in the original high-dimensional search space.
Our method is free of any low dimensional structure assumption on the function
unlike many recent high-dimensional BO methods. Optimising acquisition function
in low dimensional subspaces allows our method to obtain accurate solutions
within limited computational budget. We show that in spite of this convenience,
our algorithm remains convergent. In particular, cumulative regret of our
algorithm only grows sub-linearly with the number of iterations. More
importantly, as evident from our regret bounds, our algorithm provides a way to
trade the convergence rate with the number of subspaces used in the
optimisation. Finally, when the number of subspaces is "sufficiently large",
our algorithm's cumulative regret is at most
$\mathcal{O}^{*}(\sqrt{T\gamma_T})$ as opposed to
$\mathcal{O}^{*}(\sqrt{DT\gamma_T})$ for the GP-UCB of Srinivas et al. (2012),
reducing a crucial factor $\sqrt{D}$ where $D$ being the dimensional number of
input space.
| stat |
Variable Selection with Rigorous Uncertainty Quantification using Deep
Bayesian Neural Networks: Posterior Concentration and Bernstein-von Mises
Phenomenon | This work develops rigorous theoretical basis for the fact that deep Bayesian
neural network (BNN) is an effective tool for high-dimensional variable
selection with rigorous uncertainty quantification. We develop new Bayesian
non-parametric theorems to show that a properly configured deep BNN (1) learns
the variable importance effectively in high dimensions, and its learning rate
can sometimes "break" the curse of dimensionality. (2) BNN's uncertainty
quantification for variable importance is rigorous, in the sense that its 95%
credible intervals for variable importance indeed covers the truth 95% of the
time (i.e., the Bernstein-von Mises (BvM) phenomenon). The theoretical results
suggest a simple variable selection algorithm based on the BNN's credible
intervals. Extensive simulation confirms the theoretical findings and shows
that the proposed algorithm outperforms existing classic and
neural-network-based variable selection methods, particularly in high
dimensions.
| stat |
Active Ordinal Querying for Tuplewise Similarity Learning | Many machine learning tasks such as clustering, classification, and dataset
search benefit from embedding data points in a space where distances reflect
notions of relative similarity as perceived by humans. A common way to
construct such an embedding is to request triplet similarity queries to an
oracle, comparing two objects with respect to a reference. This work
generalizes triplet queries to tuple queries of arbitrary size that ask an
oracle to rank multiple objects against a reference, and introduces an
efficient and robust adaptive selection method called InfoTuple that uses a
novel approach to mutual information maximization. We show that the performance
of InfoTuple at various tuple sizes exceeds that of the state-of-the-art
adaptive triplet selection method on synthetic tests and new human response
datasets, and empirically demonstrate the significant gains in efficiency and
query consistency achieved by querying larger tuples instead of triplets.
| stat |
Reverse-Bayes methods: a review of recent technical advances | It is now widely accepted that the standard inferential toolkit used by the
scientific research community -- null-hypothesis significance testing (NHST) --
is not fit for purpose. Yet despite the threat posed to the scientific
enterprise, there is no agreement concerning alternative approaches. This lack
of consensus reflects long-standing issues concerning Bayesian methods, the
principal alternative to NHST. We report on recent work that builds on an
approach to inference put forward over 70 years ago to address the well-known
"Problem of Priors" in Bayesian analysis, by reversing the conventional
prior-likelihood-posterior ("forward") use of Bayes's Theorem. Such
Reverse-Bayes analysis allows priors to be deduced from the likelihood by
requiring that the posterior achieve a specified level of credibility. We
summarise the technical underpinning of this approach, and show how it opens up
new approaches to common inferential challenges, such as assessing the
credibility of scientific findings, setting them in appropriate context,
estimating the probability of successful replications, and extracting more
insight from NHST while reducing the risk of misinterpretation. We argue that
Reverse-Bayes methods have a key role to play in making Bayesian methods more
accessible and attractive to the scientific community. As a running example we
consider a recently published meta-analysis from several randomized controlled
clinical trials investigating the association between corticosteroids and
mortality in hospitalized patients with COVID-19.
| stat |
Mean and median bias reduction in generalized linear models | This paper presents an integrated framework for estimation and inference from
generalized linear models using adjusted score equations that result in mean
and median bias reduction. The framework unifies theoretical and methodological
aspects of past research on mean bias reduction and accommodates, in a natural
way, new advances on median bias reduction. General expressions for the
adjusted score functions are derived in terms of quantities that are readily
available in standard software for fitting generalized linear models. The
resulting estimating equations are solved using a unifying quasi-Fisher scoring
algorithm that is shown to be equivalent to iteratively re-weighted least
squares with appropriately adjusted working variates. Formal links between the
iterations for mean and median bias reduction are established. Core model
invariance properties are used to develop a novel mixed adjustment strategy
when the estimation of a dispersion parameter is necessary. It is also shown
how median bias reduction in multinomial logistic regression can be done using
the equivalent Poisson log-linear model. The estimates coming out from mean and
median bias reduction are found to overcome practical issues related to
infinite estimates that can occur with positive probability in generalized
linear models with multinomial or discrete responses, and can result in valid
inferences even in the presence of a high-dimensional nuisance parameter
| stat |
Identifying Critical States by the Action-Based Variance of Expected
Return | The balance of exploration and exploitation plays a crucial role in
accelerating reinforcement learning (RL). To deploy an RL agent in human
society, its explainability is also essential. However, basic RL approaches
have difficulties in deciding when to choose exploitation as well as in
extracting useful points for a brief explanation of its operation. One reason
for the difficulties is that these approaches treat all states the same way.
Here, we show that identifying critical states and treating them specially is
commonly beneficial to both problems. These critical states are the states at
which the action selection changes the potential of success and failure
substantially. We propose to identify the critical states using the variance in
the Q-function for the actions and to perform exploitation with high
probability on the identified states. These simple methods accelerate RL in a
grid world with cliffs and two baseline tasks of deep RL. Our results also
demonstrate that the identified critical states are intuitively interpretable
regarding the crucial nature of the action selection. Furthermore, our analysis
of the relationship between the timing of the identification of especially
critical states and the rapid progress of learning suggests there are a few
especially critical states that have important information for accelerating RL
rapidly.
| stat |
Time Adaptive Gaussian Model | Multivariate time series analysis is becoming an integral part of data
analysis pipelines. Understanding the individual time point connections between
covariates as well as how these connections change in time is non-trivial. To
this aim, we propose a novel method that leverages on Hidden Markov Models and
Gaussian Graphical Models -- Time Adaptive Gaussian Model (TAGM). Our model is
a generalization of state-of-the-art methods for the inference of temporal
graphical models, its formulation leverages on both aspects of these models
providing better results than current methods. In particular,it performs
pattern recognition by clustering data points in time; and, it finds
probabilistic (and possibly causal) relationships among the observed variables.
Compared to current methods for temporal network inference, it reduces the
basic assumptions while still showing good inference performances.
| stat |
Modeling event cascades using networks of additive count sequences | We propose a statistical model for networks of event count sequences built on
a cascade structure. We assume that each event triggers successor events, whose
counts follow additive probability distributions; the ensemble of counts is
given by their superposition. These assumptions allow the marginal distribution
of count sequences and the conditional distribution of event cascades to take
analytic forms. We present our model framework using Poisson and negative
binomial distributions as the building blocks. Based on this formulation, we
describe a statistical method for estimating the model parameters and event
cascades from the observed count sequences.
| stat |
Learning Asymmetric and Local Features in Multi-Dimensional Data through
Wavelets with Recursive Partitioning | Effective learning of asymmetric and local features in images and other data
observed on multi-dimensional grids is a challenging objective critical for a
wide range of image processing applications involving biomedical and natural
images. It requires methods that are sensitive to local details while fast
enough to handle massive numbers of images of ever increasing sizes. We
introduce a probabilistic model-based framework that achieves these objectives
by incorporating adaptivity into discrete wavelet transforms (DWT) through
Bayesian hierarchical modeling, thereby allowing wavelet bases to adapt to the
geometric structure of the data while maintaining the high computational
scalability of wavelet methods---linear in the sample size (e.g., the
resolution of an image). We derive a recursive representation of the Bayesian
posterior model which leads to an exact message passing algorithm to complete
learning and inference. While our framework is applicable to a range of
problems including multi-dimensional signal processing, compression, and
structural learning, we illustrate its work and evaluate its performance in the
context of image reconstruction using real images from the ImageNet database,
two widely used benchmark datasets, and a dataset from retinal optical
coherence tomography and compare its performance to state-of-the-art methods
based on basis transforms and deep learning.
| stat |
Modelling antimicrobial prescriptions in Scotland: A spatio-temporal
clustering approach | In 2016 the British government acknowledged the importance of reducing
antimicrobial prescriptions in order to avoid the long-term harmful effects of
over-prescription. Prescription needs are highly dependent on factors that have
a spatio-temporal component, such as the presence of a bacterial outbreak and
the population density. In this context, density-based clustering algorithms
are flexible tools to analyse data by searching for group structures. The case
of Scotland presents an additional challenge due to the diversity of population
densities under the area of study. We present here a spatio-temporal clustering
approach for highlighting the behaviour of general practitioners (GPs) in
Scotland. Particularly, we consider the density-based spatial clustering of
applications with noise algorithm (DBSCAN) due to its ability to include both
spatial and temporal data, as well as its flexibility to be extended with
further variables. We extend this approach into two directions. For the
temporal analysis, we use dynamic time warping to measure the dissimilarity
between warped and shifted time series. For the spatial component, we introduce
a new way of weighting spatial distances with continuous weights derived from a
KDE-based process. This makes our approach suitable for cases involving spatial
clusters with differing densities, which is a well-known issue for the original
DBSCAN. We show an improved performance compared to both the latter and the
popular k-means algorithm on simulated, as well as empirical data, presenting
evidence for the ability to cluster more elements correctly and deliver
actionable insights.
| stat |
Innovative And Additive Outlier Robust Kalman Filtering With A Robust
Particle Filter | In this paper, we propose CE-BASS, a particle mixture Kalman filter which is
robust to both innovative and additive outliers, and able to fully capture
multi-modality in the distribution of the hidden state. Furthermore, the
particle sampling approach re-samples past states, which enables CE-BASS to
handle innovative outliers which are not immediately visible in the
observations, such as trend changes. The filter is computationally efficient as
we derive new, accurate approximations to the optimal proposal distributions
for the particles. The proposed algorithm is shown to compare well with
existing approaches and is applied to both machine temperature and server data.
| stat |
Decision Making with Machine Learning and ROC Curves | The Receiver Operating Characteristic (ROC) curve is a representation of the
statistical information discovered in binary classification problems and is a
key concept in machine learning and data science. This paper studies the
statistical properties of ROC curves and its implication on model selection. We
analyze the implications of different models of incentive heterogeneity and
information asymmetry on the relation between human decisions and the ROC
curves. Our theoretical discussion is illustrated in the context of a large
data set of pregnancy outcomes and doctor diagnosis from the Pre-Pregnancy
Checkups of reproductive age couples in Henan Province provided by the Chinese
Ministry of Health.
| stat |
Spectral Non-Convex Optimization for Dimension Reduction with
Hilbert-Schmidt Independence Criterion | The Hilbert Schmidt Independence Criterion (HSIC) is a kernel dependence
measure that has applications in various aspects of machine learning.
Conveniently, the objectives of different dimensionality reduction applications
using HSIC often reduce to the same optimization problem. However, the
nonconvexity of the objective function arising from non-linear kernels poses a
serious challenge to optimization efficiency and limits the potential of
HSIC-based formulations. As a result, only linear kernels have been
computationally tractable in practice. This paper proposes a spectral-based
optimization algorithm that extends beyond the linear kernel. The algorithm
identifies a family of suitable kernels and provides the first and second-order
local guarantees when a fixed point is reached. Furthermore, we propose a
principled initialization strategy, thereby removing the need to repeat the
algorithm at random initialization points. Compared to state-of-the-art
optimization algorithms, our empirical results on real data show a run-time
improvement by as much as a factor of $10^5$ while consistently achieving lower
cost and classification/clustering errors. The implementation source code is
publicly available on https://github.com/endsley.
| stat |
Nearly Dimension-Independent Sparse Linear Bandit over Small Action
Spaces via Best Subset Selection | We consider the stochastic contextual bandit problem under the high
dimensional linear model. We focus on the case where the action space is finite
and random, with each action associated with a randomly generated contextual
covariate. This setting finds essential applications such as personalized
recommendation, online advertisement, and personalized medicine. However, it is
very challenging as we need to balance exploration and exploitation. We propose
doubly growing epochs and estimating the parameter using the best subset
selection method, which is easy to implement in practice. This approach
achieves $ \tilde{\mathcal{O}}(s\sqrt{T})$ regret with high probability, which
is nearly independent in the ``ambient'' regression model dimension $d$. We
further attain a sharper $\tilde{\mathcal{O}}(\sqrt{sT})$ regret by using the
\textsc{SupLinUCB} framework and match the minimax lower bound of
low-dimensional linear stochastic bandit problems. Finally, we conduct
extensive numerical experiments to demonstrate the applicability and robustness
of our algorithms empirically.
| stat |
MEMe: An Accurate Maximum Entropy Method for Efficient Approximations in
Large-Scale Machine Learning | Efficient approximation lies at the heart of large-scale machine learning
problems. In this paper, we propose a novel, robust maximum entropy algorithm,
which is capable of dealing with hundreds of moments and allows for
computationally efficient approximations. We showcase the usefulness of the
proposed method, its equivalence to constrained Bayesian variational inference
and demonstrate its superiority over existing approaches in two applications,
namely, fast log determinant estimation and information-theoretic Bayesian
optimisation.
| stat |
Guarantees for Tuning the Step Size using a Learning-to-Learn Approach | Learning-to-learn (using optimization algorithms to learn a new optimizer)
has successfully trained efficient optimizers in practice. This approach relies
on meta-gradient descent on a meta-objective based on the trajectory that the
optimizer generates. However, there were few theoretical guarantees on how to
avoid meta-gradient explosion/vanishing problems, or how to train an optimizer
with good generalization performance. In this paper, we study the
learning-to-learn approach on a simple problem of tuning the step size for
quadratic loss. Our results show that although there is a way to design the
meta-objective so that the meta-gradient remain polynomially bounded, computing
the meta-gradient directly using backpropagation leads to numerical issues that
look similar to gradient explosion/vanishing problems. We also characterize
when it is necessary to compute the meta-objective on a separate validation set
instead of the original training set. Finally, we verify our results
empirically and show that a similar phenomenon appears even for more
complicated learned optimizers parametrized by neural networks.
| stat |
Tensor Variable Elimination for Plated Factor Graphs | A wide class of machine learning algorithms can be reduced to variable
elimination on factor graphs. While factor graphs provide a unifying notation
for these algorithms, they do not provide a compact way to express repeated
structure when compared to plate diagrams for directed graphical models. To
exploit efficient tensor algebra in graphs with plates of variables, we
generalize undirected factor graphs to plated factor graphs and variable
elimination to a tensor variable elimination algorithm that operates directly
on plated factor graphs. Moreover, we generalize complexity bounds based on
treewidth and characterize the class of plated factor graphs for which
inference is tractable. As an application, we integrate tensor variable
elimination into the Pyro probabilistic programming language to enable exact
inference in discrete latent variable models with repeated structure. We
validate our methods with experiments on both directed and undirected graphical
models, including applications to polyphonic music modeling, animal movement
modeling, and latent sentiment analysis.
| stat |
Valid p-Values and Expectations of p-Values Revisited | A storm of favorable or critical publications regarding p-values-based
procedures has been observed in both the theoretical and applied literature. We
focus on valid definitions of p-values in the scenarios when composite null
models are in effect. A valid p-value (VpV) statistic can be used to make a
prefixed level-decision. In this context, Kolmogorov Smirnov goodness-of-fit
tests and the normal two sample problem are considered. In particular, we
examine an issue regarding the goodness-of-fit testability based on a single
observation. This article exemplifies constructions of new test procedures,
advocating practical reasons to implement VpV-based mechanisms. The VpV
framework induces an extension of the conventional expected p-value (EPV) tool
for measuring the performance of a test. Associating the EPV concept with the
receiver operating characteristic (ROC) curve methodology, a well-established
biostatistical approach, we propose a Youden index based optimality principle
to derive critical values of decision making procedures. In these terms, the
significance level alpha=0.05 can be suggested, in many situations. In light of
an ROC curve analysis, we introduce partial EPVs to characterize properties of
tests including their unbiasedness. We also provide the intrinsic relationship
between the Bayes Factor (BF) test statistic and the BF of test statistics.
Keywords: AUC; Bayes Factor; Expected p-value; Kolmogorov Smirnov tests;
Likelihood ratio; Nuisance parameters; P-value; ROC curve; Pooled data; Single
observation; Type I error rate; Youden index
| stat |
Model-assisted estimation through random forests in finite population
sampling | Surveys are used to collect data on a subset of a finite population. Most
often, the interest lies in estimating finite population parameters such as
population totals and means. In some surveys, auxiliary information is
available at the population level. This information may be incorporated in the
estimation procedures to increase their precision. Model-assisted procedures
may be based on parametric or nonparametric models. In this paper, we propose a
new class of model-assisted procedures based on random forests based on
partitions built at the population level as well as at the sample level. We
derive associated variance estimators and we establish the theoretical
properties of the proposed procedures. A model-calibration procedure that has
the ability to handle multiple survey variables is discussed. Finally, the
results of a simulation study suggest that the proposed point and estimation
procedures perform well in term of bias, efficiency and coverage in a wide
variety of settings.
| stat |
Bayesian Experimental Design for Implicit Models by Mutual Information
Neural Estimation | Implicit stochastic models, where the data-generation distribution is
intractable but sampling is possible, are ubiquitous in the natural sciences.
The models typically have free parameters that need to be inferred from data
collected in scientific experiments. A fundamental question is how to design
the experiments so that the collected data are most useful. The field of
Bayesian experimental design advocates that, ideally, we should choose designs
that maximise the mutual information (MI) between the data and the parameters.
For implicit models, however, this approach is severely hampered by the high
computational cost of computing posteriors and maximising MI, in particular
when we have more than a handful of design variables to optimise. In this
paper, we propose a new approach to Bayesian experimental design for implicit
models that leverages recent advances in neural MI estimation to deal with
these issues. We show that training a neural network to maximise a lower bound
on MI allows us to jointly determine the optimal design and the posterior.
Simulation studies illustrate that this gracefully extends Bayesian
experimental design for implicit models to higher design dimensions.
| stat |
Stratified cross-validation for unbiased and privacy-preserving
federated learning | Large-scale collections of electronic records constitute both an opportunity
for the development of more accurate prediction models and a threat for
privacy. To limit privacy exposure new privacy-enhancing techniques are
emerging such as federated learning which enables large-scale data analysis
while avoiding the centralization of records in a unique database that would
represent a critical point of failure. Although promising regarding privacy
protection, federated learning prevents using some data-cleaning algorithms
thus inducing new biases. In this work we focus on the recurrent problem of
duplicated records that, if not handled properly, may cause over-optimistic
estimations of a model's performances. We introduce and discuss stratified
cross-validation, a validation methodology that leverages stratification
techniques to prevent data leakage in federated learning settings without
relying on demanding deduplication algorithms.
| stat |
A Semi-Smooth Newton Algorithm for High-Dimensional Nonconvex Sparse
Learning | The smoothly clipped absolute deviation (SCAD) and the minimax concave
penalty (MCP) penalized regression models are two important and widely used
nonconvex sparse learning tools that can handle variable selection and
parameter estimation simultaneously, and thus have potential applications in
various fields such as mining biological data in high-throughput biomedical
studies. Theoretically, these two models enjoy the oracle property even in the
high-dimensional settings, where the number of predictors $p$ may be much
larger than the number of observations $n$. However, numerically, it is quite
challenging to develop fast and stable algorithms due to their non-convexity
and non-smoothness. In this paper we develop a fast algorithm for SCAD and MCP
penalized learning problems. First, we show that the global minimizers of both
models are roots of the nonsmooth equations. Then, a semi-smooth Newton (SSN)
algorithm is employed to solve the equations. We prove that the SSN algorithm
converges locally and superlinearly to the Karush-Kuhn-Tucker (KKT) points.
Computational complexity analysis shows that the cost of the SSN algorithm per
iteration is $O(np)$. Combined with the warm-start technique, the SSN algorithm
can be very efficient and accurate. Simulation studies and a real data example
suggest that our SSN algorithm, with comparable solution accuracy with the
coordinate descent (CD) and the difference of convex (DC) proximal Newton
algorithms, is more computationally efficient.
| stat |
Testing Independence under Biased Sampling | Testing for association or dependence between pairs of random variables is a
fundamental problem in statistics. In some applications, data are subject to
selection bias that causes dependence between observations even when it is
absent from the population. An important example is truncation models, in which
observed pairs are restricted to a specific subset of the X-Y plane. Standard
tests for independence are not suitable in such cases, and alternative tests
that take the selection bias into account are required. To deal with this
issue, we generalize the notion of quasi-independence with respect to the
sampling mechanism, and study the problem of detecting any deviations from it.
We develop two test statistics motivated by the classic Hoeffding's statistic,
and use two approaches to compute their distribution under the null: (i) a
bootstrap-based approach, and (ii) a permutation-test with non-uniform
probability of permutations, sampled using either MCMC or importance sampling
with various proposal distributions. We show that our tests can tackle cases
where the biased sampling mechanism is estimated from the data, with an
important application to the case of censoring with truncation. We prove the
validity of the tests, and show, using simulations, that they perform well for
important special cases of the problem and improve power compared to competing
methods. The tests are applied to four datasets, two that are subject to
truncation, with and without censoring, and two to positive bias mechanisms
related to length bias.
| stat |
Convex Nonparanormal Regression | Quantifying uncertainty in predictions or, more generally, estimating the
posterior conditional distribution, is a core challenge in machine learning and
statistics. We introduce Convex Nonparanormal Regression (CNR), a conditional
nonparanormal approach for coping with this task. CNR involves a convex
optimization of a posterior defined via a rich dictionary of pre-defined non
linear transformations on Gaussians. It can fit an arbitrary conditional
distribution, including multimodal and non-symmetric posteriors. For the
special but powerful case of a piecewise linear dictionary, we provide a closed
form of the posterior mean which can be used for point-wise predictions.
Finally, we demonstrate the advantages of CNR over classical competitors using
synthetic and real world data.
| stat |
Towards reliable projections of global mean surface temperature | Quantifying the risk of global warming exceeding critical targets such as 2.0
K requires reliable projections of uncertainty as well as best estimates of
Global Mean Surface Temperature (GMST). However, uncertainty bands on GMST
projections are often calculated heuristically and have several potential
shortcomings. In particular, the uncertainty bands shown in IPCC plume
projections of GMST are based on the distribution of GMST anomalies from
climate model runs and so are strongly determined by model characteristics with
little influence from observations of the real-world. Physically motivated
time-series approaches are proposed based on fitting energy balance models
(EBMs) to climate model outputs and observations in order to constrain future
projections. It is shown that EBMs fitted to one forcing scenario will not
produce reliable projections when different forcing scenarios are applied. The
errors in the EBM projections can be interpreted as arising due to a
discrepancy in the effective forcing felt by the model. A simple time-series
approach to correcting the projections is proposed based on learning the
evolution of the forcing discrepancy so that it can be projected into the
future. These approaches give reliable projections of GMST when tested in a
perfect model setting, and when applied to observations lead to well
constrained projections with lower mean warming and narrower projection bands
than previous estimates. Despite the reduced uncertainty, the lower warming
leads to a greatly reduced probability of exceeding the 2.0 K warming target.
| stat |
Faster MCMC for Gaussian Latent Position Network Models | Latent position network models are a versatile tool in network science;
applications include clustering entities, controlling for causal confounders,
and defining priors over unobserved graphs. Estimating each node's latent
position is typically framed as a Bayesian inference problem, with Metropolis
within Gibbs being the most popular tool for approximating the posterior
distribution. However, it is well-known that Metropolis within Gibbs is
inefficient for large networks; the acceptance ratios are expensive to compute,
and the resultant posterior draws are highly correlated. In this article, we
propose an alternative Markov chain Monte Carlo strategy---defined using a
combination of split Hamiltonian Monte Carlo and Firefly Monte Carlo---that
leverages the posterior distribution's functional form for more efficient
posterior computation. We demonstrate that these strategies outperform
Metropolis within Gibbs and other algorithms on synthetic networks, as well as
on real information-sharing networks of teachers and staff in a school
district.
| stat |
Practical Calculation of Gittins Indices for Multi-armed Bandits | Gittins indices provide an optimal solution to the classical multi-armed
bandit problem. An obstacle to their use has been the common perception that
their computation is very difficult. This paper demonstrates an accessible
general methodology for the calculating Gittins indices for the multi-armed
bandit with a detailed study on the cases of Bernoulli and Gaussian rewards.
With accompanying easy-to-use open source software, this work removes
computation as a barrier to using Gittins indices in these commonly found
settings.
| stat |
Modern Monte Carlo Methods for Efficient Uncertainty Quantification and
Propagation: A Survey | Uncertainty quantification (UQ) includes the characterization, integration,
and propagation of uncertainties that result from stochastic variations and a
lack of knowledge or data in the natural world. Monte Carlo (MC) method is a
sampling-based approach that has widely used for quantification and propagation
of uncertainties. However, the standard MC method is often time-consuming if
the simulation-based model is computationally intensive. This article gives an
overview of modern MC methods to address the existing challenges of the
standard MC in the context of UQ. Specifically, multilevel Monte Carlo (MLMC)
extending the concept of control variates achieves a significant reduction of
the computational cost by performing most evaluations with low accuracy and
corresponding low cost, and relatively few evaluations at high accuracy and
correspondingly high cost. Multifidelity Monte Carlo (MFMC) accelerates the
convergence of standard Monte Carlo by generalizing the control variates with
different models having varying fidelities and varying computational costs.
Multimodel Monte Carlo method (MMMC), having a different setting of MLMC and
MFMC, aims to address the issue of uncertainty quantification and propagation
when data for characterizing probability distributions are limited. Multimodel
inference combined with importance sampling is proposed for quantifying and
efficiently propagating the uncertainties resulting from small datasets. All of
these three modern MC methods achieve a significant improvement of
computational efficiency for probabilistic UQ, particularly uncertainty
propagation. An algorithm summary and the corresponding code implementation are
provided for each of the modern Monte Carlo methods. The extension and
application of these methods are discussed in detail.
| stat |
Fast estimation of a convolution type model for the intensity of spatial
point processes | Estimating the first-order intensity function in point pattern analysis is an
important problem, and it has been approached so far from different
perspectives: parametrically, semiparametrically or nonparametrically. Our
approach is close to a semiparametric one. Motivated by eye-movement data, we
introduce a convolution type model where the log-intensity is modelled as the
convolution of a function $\beta(\cdot)$, to be estimated, and a single spatial
covariate (the image an individual is looking at for eye-movement data). Based
on a Fourier series expansion, we show that the proposed model is related to
the log-linear model with infinite number of coefficients, which correspond to
the spectral decomposition of $\beta(\cdot)$. After truncation, we estimate
these coefficients through a penalized Poisson likelihood and prove infill
asymptotic results for a large class of spatial point processes. We illustrate
the efficiency of the proposed methodology on simulated data and real data.
| stat |
Generic probabilistic modelling and non-homogeneity issues for the UK
epidemic of COVID-19 | Coronavirus COVID-19 spreads through the population mostly based on social
contact. To gauge the potential for widespread contagion, to cope with
associated uncertainty and to inform its mitigation, more accurate and robust
modelling is centrally important for policy making. We provide a flexible
modelling approach that increases the accuracy with which insights can be made.
We use this to analyse different scenarios relevant to the COVID-19 situation
in the UK. We present a stochastic model that captures the inherently
probabilistic nature of contagion between population members. The computational
nature of our model means that spatial constraints (e.g., communities and
regions), the susceptibility of different age groups and other factors such as
medical pre-histories can be incorporated with ease. We analyse different
possible scenarios of the COVID-19 situation in the UK. Our model is robust to
small changes in the parameters and is flexible in being able to deal with
different scenarios.
This approach goes beyond the convention of representing the spread of an
epidemic through a fixed cycle of susceptibility, infection and recovery (SIR).
It is important to emphasise that standard SIR-type models, unlike our model,
are not flexible enough and are also not stochastic and hence should be used
with extreme caution. Our model allows both heterogeneity and inherent
uncertainty to be incorporated. Due to the scarcity of verified data, we draw
insights by calibrating our model using parameters from other relevant sources,
including agreement on average (mean field) with parameters in SIR-based
models.
| stat |
Faking Fairness via Stealthily Biased Sampling | Auditing fairness of decision-makers is now in high demand. To respond to
this social demand, several fairness auditing tools have been developed. The
focus of this study is to raise an awareness of the risk of malicious
decision-makers who fake fairness by abusing the auditing tools and thereby
deceiving the social communities. The question is whether such a fraud of the
decision-maker is detectable so that the society can avoid the risk of fake
fairness. In this study, we answer this question negatively. We specifically
put our focus on a situation where the decision-maker publishes a benchmark
dataset as the evidence of his/her fairness and attempts to deceive a person
who uses an auditing tool that computes a fairness metric. To assess the
(un)detectability of the fraud, we explicitly construct an algorithm, the
stealthily biased sampling, that can deliberately construct an evil benchmark
dataset via subsampling. We show that the fraud made by the stealthily based
sampling is indeed difficult to detect both theoretically and empirically.
| stat |
Causal inference of hazard ratio based on propensity score matching | Propensity score matching is commonly used to draw causal inference from
observational survival data. However, there is no gold standard approach to
analyze survival data after propensity score matching, and variance estimation
after matching is open to debate. We derive the statistical properties of the
propensity score matching estimator of the marginal causal hazard ratio based
on matching with replacement and a fixed number of matches. We also propose a
double-resampling technique for variance estimation that takes into account the
uncertainty due to propensity score estimation prior to matching.
| stat |
Best-scored Random Forest Density Estimation | This paper presents a brand new nonparametric density estimation strategy
named the best-scored random forest density estimation whose effectiveness is
supported by both solid theoretical analysis and significant experimental
performance. The terminology best-scored stands for selecting one density tree
with the best estimation performance out of a certain number of purely random
density tree candidates and we then name the selected one the best-scored
random density tree. In this manner, the ensemble of these selected trees that
is the best-scored random density forest can achieve even better estimation
results than simply integrating trees without selection. From the theoretical
perspective, by decomposing the error term into two, we are able to carry out
the following analysis: First of all, we establish the consistency of the
best-scored random density trees under $L_1$-norm. Secondly, we provide the
convergence rates of them under $L_1$-norm concerning with three different tail
assumptions, respectively. Thirdly, the convergence rates under
$L_{\infty}$-norm is presented. Last but not least, we also achieve the above
convergence rates analysis for the best-scored random density forest. When
conducting comparative experiments with other state-of-the-art density
estimation approaches on both synthetic and real data sets, it turns out that
our algorithm has not only significant advantages in terms of estimation
accuracy over other methods, but also stronger resistance to the curse of
dimensionality.
| stat |
TSEC: a framework for online experimentation under experimental
constraints | Thompson sampling is a popular algorithm for solving multi-armed bandit
problems, and has been applied in a wide range of applications, from website
design to portfolio optimization. In such applications, however, the number of
choices (or arms) $N$ can be large, and the data needed to make adaptive
decisions require expensive experimentation. One is then faced with the
constraint of experimenting on only a small subset of $K \ll N$ arms within
each time period, which poses a problem for traditional Thompson sampling. We
propose a new Thompson Sampling under Experimental Constraints (TSEC) method,
which addresses this so-called "arm budget constraint". TSEC makes use of a
Bayesian interaction model with effect hierarchy priors, to model correlations
between rewards on different arms. This fitted model is then integrated within
Thompson sampling, to jointly identify a good subset of arms for
experimentation and to allocate resources over these arms. We demonstrate the
effectiveness of TSEC in two problems with arm budget constraints. The first is
a simulated website optimization study, where TSEC shows noticeable
improvements over industry benchmarks. The second is a portfolio optimization
application on industry-based exchange-traded funds, where TSEC provides more
consistent and greater wealth accumulation over standard investment strategies.
| stat |
Balancing the Tradeoff Between Clustering Value and Interpretability | Graph clustering groups entities -- the vertices of a graph -- based on their
similarity, typically using a complex distance function over a large number of
features. Successful integration of clustering approaches in automated
decision-support systems hinges on the interpretability of the resulting
clusters. This paper addresses the problem of generating interpretable
clusters, given features of interest that signify interpretability to an
end-user, by optimizing interpretability in addition to common clustering
objectives. We propose a $\beta$-interpretable clustering algorithm that
ensures that at least $\beta$ fraction of nodes in each cluster share the same
feature value. The tunable parameter $\beta$ is user-specified. We also present
a more efficient algorithm for scenarios with $\beta\!=\!1$ and analyze the
theoretical guarantees of the two algorithms. Finally, we empirically
demonstrate the benefits of our approaches in generating interpretable clusters
using four real-world datasets. The interpretability of the clusters is
complemented by generating simple explanations denoting the feature values of
the nodes in the clusters, using frequent pattern mining.
| stat |
Posterior inference unchained with EL_2O | Statistical inference of analytically non-tractable posteriors is a difficult
problem because of marginalization of correlated variables and stochastic
methods such as MCMC and VI are commonly used. We argue that stochastic KL
divergence minimization used by MCMC and VI is noisy, and we propose instead
EL_2O, expectation optimization of L_2 distance squared between the approximate
log posterior q and the un-normalized log posterior of p. When sampling from q
the solutions agree with stochastic KL divergence minimization based VI in the
large sample limit, however EL_2O method is free of sampling noise, has better
optimization properties, and requires only as many sample evaluations as the
number of parameters we are optimizing if q covers p. As a consequence,
increasing the expressivity of q improves both the quality of results and the
convergence rate, allowing EL_2O to approach exact inference. Use of automatic
differentiation methods enables us to develop Hessian, gradient and gradient
free versions of the method, which can determine M(M+2)/2+1, M+1 and 1
parameter(s) of q with a single sample, respectively. EL_2O provides a reliable
estimate of the quality of the approximating posterior, and converges rapidly
on full rank gaussian approximation for q and extensions beyond it, such as
nonlinear transformations and gaussian mixtures. These can handle general
posteriors, while still allowing fast analytic marginalizations. We test it on
several examples, including a realistic 13 dimensional galaxy clustering
analysis, showing that it is several orders of magnitude faster than MCMC,
while giving smooth and accurate non-gaussian posteriors, often requiring a few
to a few dozen of iterations only.
| stat |
Stochastic Approximation EM for Exploratory Item Factor Analysis | The stochastic approximation EM algorithm (SAEM) is described for the
estimation of item and person parameters given test data coded as dichotomous
or ordinal variables. The method hinges upon the eigenanalysis of missing
variables sampled as augmented data; the augmented data approach was introduced
by Albert's seminal work applying Gibbs sampling to Item Response Theory in
1992. Similar to maximum likelihood factor analysis, the factor structure in
this Bayesian approach depends only on sufficient statistics, which are
computed from the missing latent data. A second feature of the SAEM algorithm
is the use of the Robbins-Monro procedure for establishing convergence.
Contrary to Expectation Maximization methods where costly integrals must be
calculated, this method is well-suited for highly multidimensional data, and an
annealing method is implemented to prevent convergence to a local maximum
likelihood. Multiple calculations of errors applied within this framework of
Markov Chain Monte Carlo are presented to delineate the uncertainty of
parameter estimates. Given the nature of EFA (exploratory factor analysis), an
algorithm is formalized leveraging the Tracy-Widom distribution for the
retention of factors extracted from an eigenanalysis of the sufficient
statistic of the covariance of the augmented data matrix. Simulation conditions
of dichotomous and polytomous data, from one to ten dimensions of factor
loadings, are used to assess statistical accuracy and to gauge computational
time of the EFA approach of this IRT-specific implementation of the SAEM
algorithm. Finally, three applications of this methodology are also reported
that demonstrate the effectiveness of the method for enabling timely analyses
as well as substantive interpretations when this method is applied to real
data.
| stat |
The Benefits of Probability-Proportional-to-Size Sampling in
Cluster-Randomized Experiments | In a cluster-randomized experiment, treatment is assigned to clusters of
individual units of interest--households, classrooms, villages, etc.--instead
of the units themselves. The number of clusters sampled and the number of units
sampled within each cluster is typically restricted by a budget constraint.
Previous analysis of cluster randomized experiments under the Neyman-Rubin
potential outcomes model of response have assumed a simple random sample of
clusters. Estimators of the population average treatment effect (PATE) under
this assumption are often either biased or not invariant to location shifts of
potential outcomes. We demonstrate that, by sampling clusters with probability
proportional to the number of units within a cluster, the Horvitz-Thompson
estimator (HT) is invariant to location shifts and unbiasedly estimates PATE.
We derive standard errors of HT and discuss how to estimate these standard
errors. We also show that results hold for stratified random samples when
samples are drawn proportionally to cluster size within each stratum. We
demonstrate the efficacy of this sampling scheme using a simulation based on
data from an experiment measuring the efficacy of the National Solidarity
Programme in Afghanistan.
| stat |
Causal Inference on Non-linear Spaces: Distribution Functions and Beyond | Understanding causal relationships is one of the most important goals of
modern science. So far, the causal inference literature has focused almost
exclusively on outcomes coming from a linear space, most commonly the Euclidean
space. However, it is increasingly common that complex datasets collected
through electronic sources, such as wearable devices and medical imaging,
cannot be represented as data points from linear spaces. In this paper, we
present a formal definition of causal effects for outcomes from non-linear
spaces, with a focus on the Wasserstein space of cumulative distribution
functions. We develop doubly robust estimators and associated asymptotic theory
for these causal effects. Our framework extends to outcomes from certain
Riemannian manifolds. As an illustration, we use our framework to quantify the
causal effect of marriage on physical activity patterns using wearable device
data collected through the National Health and Nutrition Examination Survey.
| stat |
Meta-Cal: Well-controlled Post-hoc Calibration by Ranking | In many applications, it is desirable that a classifier not only makes
accurate predictions, but also outputs calibrated probabilities. However, many
existing classifiers, especially deep neural network classifiers, tend not to
be calibrated. Post-hoc calibration is a technique to recalibrate a model, and
its goal is to learn a calibration map. Existing approaches mostly focus on
constructing calibration maps with low calibration errors. Contrary to these
methods, we study post-hoc calibration for multi-class classification under
constraints, as a calibrator with a low calibration error does not necessarily
mean it is useful in practice. In this paper, we introduce two practical
constraints to be taken into consideration. We then present Meta-Cal, which is
built from a base calibrator and a ranking model. Under some mild assumptions,
two high-probability bounds are given with respect to these constraints.
Empirical results on CIFAR-10, CIFAR-100 and ImageNet and a range of popular
network architectures show our proposed method significantly outperforms the
current state of the art for post-hoc multi-class classification calibration.
| stat |
Exact Asymptotics for Learning Tree-Structured Graphical Models with
Side Information: Noiseless and Noisy Samples | Given side information that an Ising tree-structured graphical model is
homogeneous and has no external field, we derive the exact asymptotics of
learning its structure from independently drawn samples. Our results, which
leverage the use of probabilistic tools from the theory of strong large
deviations, refine the large deviation (error exponents) results of Tan,
Anandkumar, Tong, and Willsky [IEEE Trans. on Inform. Th., 57(3):1714--1735,
2011] and strictly improve those of Bresler and Karzand [Ann. Statist., 2020].
In addition, we extend our results to the scenario in which the samples are
observed in random noise. In this case, we show that they strictly improve on
the recent results of Nikolakakis, Kalogerias, and Sarwate [Proc. AISTATS,
1771--1782, 2019]. Our theoretical results demonstrate keen agreement with
experimental results for sample sizes as small as that in the hundreds.
| stat |
Recent advances in directional statistics | Mainstream statistical methodology is generally applicable to data observed
in Euclidean space. There are, however, numerous contexts of considerable
scientific interest in which the natural supports for the data under
consideration are Riemannian manifolds like the unit circle, torus, sphere and
their extensions. Typically, such data can be represented using one or more
directions, and directional statistics is the branch of statistics that deals
with their analysis. In this paper we provide a review of the many recent
developments in the field since the publication of Mardia and Jupp (1999),
still the most comprehensive text on directional statistics. Many of those
developments have been stimulated by interesting applications in fields as
diverse as astronomy, medicine, genetics, neurology, aeronautics, acoustics,
image analysis, text mining, environmetrics, and machine learning. We begin by
considering developments for the exploratory analysis of directional data
before progressing to distributional models, general approaches to inference,
hypothesis testing, regression, nonparametric curve estimation, methods for
dimension reduction, classification and clustering, and the modelling of time
series, spatial and spatio-temporal data. An overview of currently available
software for analysing directional data is also provided, and potential future
developments discussed.
| stat |
Meta Cyclical Annealing Schedule: A Simple Approach to Avoiding
Meta-Amortization Error | The ability to learn new concepts with small amounts of data is a crucial
aspect of intelligence that has proven challenging for deep learning methods.
Meta-learning for few-shot learning offers a potential solution to this
problem: by learning to learn across data from many previous tasks, few-shot
learning algorithms can discover the structure among tasks to enable fast
learning of new tasks. However, a critical challenge in few-shot learning is
task ambiguity: even when a powerful prior can be meta-learned from a large
number of prior tasks, a small dataset for a new task can simply be very
ambiguous to acquire a single model for that task. The Bayesian meta-learning
models can naturally resolve this problem by putting a sophisticated prior
distribution and let the posterior well regularized through Bayesian decision
theory. However, currently known Bayesian meta-learning procedures such as
VERSA suffer from the so-called {\it information preference problem}, that is,
the posterior distribution is degenerated to one point and is far from the
exact one. To address this challenge, we design a novel meta-regularization
objective using {\it cyclical annealing schedule} and {\it maximum mean
discrepancy} (MMD) criterion. The cyclical annealing schedule is quite
effective at avoiding such degenerate solutions. This procedure includes a
difficult KL-divergence estimation, but we resolve the issue by employing MMD
instead of KL-divergence. The experimental results show that our approach
substantially outperforms standard meta-learning algorithms.
| stat |
The Micro-Randomized Trial for Developing Digital Interventions: Data
Analysis Methods | Although there is much excitement surrounding the use of mobile and wearable
technology for the purposes of delivering interventions as people go through
their day-to-day lives, data analysis methods for constructing and optimizing
digital interventions lag behind. Here, we elucidate data analysis methods for
primary and secondary analyses of micro-randomized trials (MRTs), an
experimental design to optimize digital just-in-time adaptive interventions. We
provide a definition of causal "excursion" effects suitable for use in digital
intervention development. We introduce the weighted and centered least-squares
(WCLS) estimator which provides consistent causal excursion effect estimators
for digital interventions from MRT data. We describe how the WCLS estimator
along with associated test statistics can be obtained using standard
statistical software such as SAS or R. Throughout we use HeartSteps, an MRT
designed to increase physical activity among sedentary individuals, to
illustrate potential primary and secondary analyses.
| stat |
Combining heterogeneous subgroups with graph-structured variable
selection priors for Cox regression | Important objectives in cancer research are the prediction of a patient's
risk based on molecular measurements such as gene expression data and the
identification of new prognostic biomarkers (e.g. genes). In clinical practice,
this is often challenging because patient cohorts are typically small and can
be heterogeneous. In classical subgroup analysis, a separate prediction model
is fitted using only the data of one specific cohort. However, this can lead to
a loss of power when the sample size is small. Simple pooling of all cohorts,
on the other hand, can lead to biased results, especially when the cohorts are
heterogeneous. For this situation, we propose a new Bayesian approach suitable
for continuous molecular measurements and survival outcome that identifies the
important predictors and provides a separate risk prediction model for each
cohort. It allows sharing information between cohorts to increase power by
assuming a graph linking predictors within and across different cohorts. The
graph helps to identify pathways of functionally related genes and genes that
are simultaneously prognostic in different cohorts. Results demonstrate that
our proposed approach is superior to the standard approaches in terms of
prediction performance and increased power in variable selection when the
sample size is small.
| stat |
Optimal post-selection inference for sparse signals: a nonparametric
empirical-Bayes approach | Many recently developed Bayesian methods have focused on sparse signal
detection. However, much less work has been done addressing the natural
follow-up question: how to make valid inferences for the magnitude of those
signals after selection. Ordinary Bayesian credible intervals suffer from
selection bias, owing to the fact that the target of inference is chosen
adaptively. Existing Bayesian approaches for correcting this bias produce
credible intervals with poor frequentist properties, while existing frequentist
approaches require sacrificing the benefits of shrinkage typical in Bayesian
methods, resulting in confidence intervals that are needlessly wide. We address
this gap by proposing a nonparametric empirical-Bayes approach for constructing
optimal selection-adjusted confidence sets. Our method produces confidence sets
that are as short as possible on average, while both adjusting for selection
and maintaining exact frequentist coverage uniformly over the parameter space.
Our main theoretical result establishes an important consistency property of
our procedure: that under mild conditions, it asymptotically converges to the
results of an oracle-Bayes analysis in which the prior distribution of signal
sizes is known exactly. Across a series of examples, the method outperforms
existing frequentist techniques for post-selection inference, producing
confidence sets that are notably shorter but with the same coverage guarantee.
| stat |
How to iron out rough landscapes and get optimal performances: Averaged
Gradient Descent and its application to tensor PCA | In many high-dimensional estimation problems the main task consists in
minimizing a cost function, which is often strongly non-convex when scanned in
the space of parameters to be estimated. A standard solution to flatten the
corresponding rough landscape consists in summing the losses associated to
different data points and obtain a smoother empirical risk. Here we propose a
complementary method that works for a single data point. The main idea is that
a large amount of the roughness is uncorrelated in different parts of the
landscape. One can then substantially reduce the noise by evaluating an
empirical average of the gradient obtained as a sum over many random
independent positions in the space of parameters to be optimized. We present an
algorithm, called Averaged Gradient Descent, based on this idea and we apply it
to tensor PCA, which is a very hard estimation problem. We show that Averaged
Gradient Descent over-performs physical algorithms such as gradient descent and
approximate message passing and matches the best algorithmic thresholds known
so far, obtained by tensor unfolding and methods based on sum-of-squares.
| stat |
Multivariate hierarchical analysis of car crashes data considering a
spatial network lattice | Road traffic casualties represent a hidden global epidemic, demanding
evidence-based interventions. This paper demonstrates a network lattice
approach for identifying road segments of particular concern, based on a case
study of a major city (Leeds, UK), in which 5862 crashes of different
severities were recorded over an eight-year period (2011-2018). We consider a
family of Bayesian hierarchical models that include spatially structured and
unstructured random effects, to capture the dependencies between the severity
levels. Results highlight roads that are more prone to collisions, relative to
estimated traffic volumes, in the North of the city. We analyse the Modifiable
Areal Unit Problem (MAUP), proposing a novel procedure to investigate the
presence of MAUP on a network lattice. We conclude that our methods enable a
reliable estimation of road safety levels to help identify hotspots on the road
network and to inform effective local interventions.
| stat |
Learning Whenever Learning is Possible: Universal Learning under General
Stochastic Processes | This work initiates a general study of learning and generalization without
the i.i.d. assumption, starting from first principles. While the traditional
approach to statistical learning theory typically relies on standard
assumptions from probability theory (e.g., i.i.d. or stationary ergodic), in
this work we are interested in developing a theory of learning based only on
the most fundamental and necessary assumptions implicit in the requirements of
the learning problem itself. We specifically study universally consistent
function learning, where the objective is to obtain low long-run average loss
for any target function, when the data follow a given stochastic process. We
are then interested in the question of whether there exist learning rules
guaranteed to be universally consistent given only the assumption that
universally consistent learning is possible for the given data process. The
reasoning that motivates this criterion emanates from a kind of optimist's
decision theory, and so we refer to such learning rules as being optimistically
universal. We study this question in three natural learning settings:
inductive, self-adaptive, and online. Remarkably, as our strongest positive
result, we find that optimistically universal learning rules do indeed exist in
the self-adaptive learning setting. Establishing this fact requires us to
develop new approaches to the design of learning algorithms. Along the way, we
also identify concise characterizations of the family of processes under which
universally consistent learning is possible in the inductive and self-adaptive
settings. We additionally pose a number of enticing open problems, particularly
for the online learning setting.
| stat |
Analyzing Basket Trials under Multisource Exchangeability Assumptions | Basket designs are prospective clinical trials that are devised with the
hypothesis that the presence of selected molecular features determine a
patient's subsequent response to a particular "targeted" treatment strategy.
Basket trials are designed to enroll multiple clinical subpopulations to which
it is assumed that the therapy in question offers beneficial efficacy in the
presence of the targeted molecular profile. The treatment, however, may not
offer acceptable efficacy to all subpopulations enrolled. Moreover, for rare
disease settings, such as oncology wherein these trials have become popular,
marginal measures of statistical evidence are difficult to interpret for
sparsely enrolled subpopulations. Consequently, basket trials pose challenges
to the traditional paradigm for trial design, which assumes inter-patient
exchangeability. The R-package \pkg{basket} facilitates the analysis of basket
trials by implementing multi-source exchangeability models. By evaluating all
possible pairwise exchangeability relationships, this hierarchical modeling
framework facilitates Bayesian posterior shrinkage among a collection of
discrete and pre-specified subpopulations. Analysis functions are provided to
implement posterior inference of the response rates and all possible
exchangeability relationships between subpopulations. In addition, the package
can identify "poolable" subsets of and report their response characteristics.
The functionality of the package is demonstrated using data from an oncology
study with subpopulations defined by tumor histology.
| stat |
Learning the piece-wise constant graph structure of a varying Ising
model | This work focuses on the estimation of multiple change-points in a
time-varying Ising model that evolves piece-wise constantly. The aim is to
identify both the moments at which significant changes occur in the Ising
model, as well as the underlying graph structures. For this purpose, we propose
to estimate the neighborhood of each node by maximizing a penalized version of
its conditional log-likelihood. The objective of the penalization is twofold:
it imposes sparsity in the learned graphs and, thanks to a fused-type penalty,
it also enforces them to evolve piece-wise constantly. Using few assumptions,
we provide two change-points consistency theorems. Those are the first in the
context of unknown number of change-points detection in time-varying Ising
model. Finally, experimental results on several synthetic datasets and a
real-world dataset demonstrate the performance of our method.
| stat |
Space Filling Split Plot Design using Fast Flexible Filling | In this article, an adaption of an algorithm for the creation of experimental
designs by Lekivetz and Jones (2015) is suggested, dealing with constraints
around randomization. Split-plot design of experiments is used, when the levels
of some factors cannot be modified as easily as others. While most split-plot
designs deal in the context of I-optimal or D-optimal designs for continuous
response outputs, a space filling design strategy is suggested in here. The
proposed designs are evaluated based on different design criteria, as well as
an analytical example.
| stat |
A Non-Gaussian Spatio-Temporal Model for Daily Wind Speeds Based on a
Multivariate Skew-t Distribution | Facing increasing domestic energy consumption from population growth and
industrialization, Saudi Arabia is aiming to reduce its reliance on fossil
fuels and to broaden its energy mix by expanding investment in renewable energy
sources, including wind energy. A preliminary task in the development of wind
energy infrastructure is the assessment of wind energy potential, a key aspect
of which is the characterization of its spatio-temporal behavior. In this study
we examine the impact of internal climate variability on seasonal wind power
density fluctuations over Saudi Arabia using 30 simulations from the Large
Ensemble Project (LENS) developed at the National Center for Atmospheric
Research. Furthermore, a spatio-temporal model for daily wind speed is proposed
with neighbor-based cross-temporal dependence, and a multivariate skew-t
distribution to capture the spatial patterns of higher order moments. The model
can be used to generate synthetic time series over the entire spatial domain
that adequately reproduce the internal variability of the LENS dataset.
| stat |
Fast Rates for Contextual Linear Optimization | Incorporating side observations in decision making can reduce uncertainty and
boost performance, but it also requires we tackle a potentially complex
predictive relationship. While one may use off-the-shelf machine learning
methods to separately learn a predictive model and plug it in, a variety of
recent methods instead integrate estimation and optimization by fitting the
model to directly optimize downstream decision performance. Surprisingly, in
the case of contextual linear optimization, we show that the naive plug-in
approach actually achieves regret convergence rates that are significantly
faster than methods that directly optimize downstream decision performance. We
show this by leveraging the fact that specific problem instances do not have
arbitrarily bad near-dual-degeneracy. While there are other pros and cons to
consider as we discuss and illustrate numerically, our results highlight a
nuanced landscape for the enterprise to integrate estimation and optimization.
Our results are overall positive for practice: predictive models are easy and
fast to train using existing tools, simple to interpret, and, as we show, lead
to decisions that perform very well.
| stat |
Causal Autoregressive Flows | Two apparently unrelated fields -- normalizing flows and causality -- have
recently received considerable attention in the machine learning community. In
this work, we highlight an intrinsic correspondence between a simple family of
autoregressive normalizing flows and identifiable causal models. We exploit the
fact that autoregressive flow architectures define an ordering over variables,
analogous to a causal ordering, to show that they are well-suited to performing
a range of causal inference tasks, ranging from causal discovery to making
interventional and counterfactual predictions. First, we show that causal
models derived from both affine and additive autoregressive flows with fixed
orderings over variables are identifiable, i.e. the true direction of causal
influence can be recovered. This provides a generalization of the additive
noise model well-known in causal discovery. Second, we derive a bivariate
measure of causal direction based on likelihood ratios, leveraging the fact
that flow models can estimate normalized log-densities of data. Third, we
demonstrate that flows naturally allow for direct evaluation of both
interventional and counterfactual queries, the latter case being possible due
to the invertible nature of flows. Finally, throughout a series of experiments
on synthetic and real data, the proposed method is shown to outperform current
approaches for causal discovery as well as making accurate interventional and
counterfactual predictions.
| stat |
Rapid Bayesian inference for expensive stochastic models | Almost all fields of science rely upon statistical inference to estimate
unknown parameters in theoretical and computational models. While the
performance of modern computer hardware continues to grow, the computational
requirements for the simulation of models are growing even faster. This is
largely due to the increase in model complexity, often including stochastic
dynamics, that is necessary to describe and characterize phenomena observed
using modern, high resolution, experimental techniques. Such models are rarely
analytically tractable, meaning that extremely large numbers of stochastic
simulations are required for parameter inference. In such cases, parameter
inference can be practically impossible. In this work, we present new
computational Bayesian techniques that accelerate inference for expensive
stochastic models by using computationally inexpensive approximations to inform
feasible regions in parameter space, and through learning transforms that
adjust the biased approximate inferences to closer represent the correct
inferences under the expensive stochastic model. Using topical examples from
ecology and cell biology, we demonstrate a speed improvement of an order of
magnitude without any loss in accuracy. This represents a substantial
improvement over current state-of-the-art methods for Bayesian computations
when appropriate model approximations are available.
| stat |
Debiased Inverse Propensity Score Weighting for Estimation of Average
Treatment Effects with High-Dimensional Confounders | We consider estimation of average treatment effects given observational data
with high-dimensional pretreatment variables. Existing methods for this problem
typically assume some form of sparsity for the regression functions. In this
work, we introduce a debiased inverse propensity score weighting (DIPW) scheme
for average treatment effect estimation that delivers $\sqrt{n}$-consistent
estimates of the average treatment effect when the propensity score follows a
sparse logistic regression model; the regression functions are permitted to be
arbitrarily complex. Our theoretical results quantify the price to pay for
permitting the regression functions to be unestimable, which shows up as an
inflation of the variance of the estimator compared to the semiparametric
efficient variance by at most O(1) under mild conditions. Given the lack of
assumptions on the regression functions, averages of transformed responses
under each treatment may also be estimated at the $\sqrt{n}$ rate, and so for
example, the variances of the potential outcomes may be estimated. We show how
confidence intervals centred on our estimates may be constructed, and also
discuss an extension of the method to estimating projections of the
heterogeneous treatment effect function.
| stat |
Variational Tracking and Prediction with Generative Disentangled
State-Space Models | We address tracking and prediction of multiple moving objects in visual data
streams as inference and sampling in a disentangled latent state-space model.
By encoding objects separately and including explicit position information in
the latent state space, we perform tracking via amortized variational Bayesian
inference of the respective latent positions. Inference is implemented in a
modular neural framework tailored towards our disentangled latent space.
Generative and inference model are jointly learned from observations only.
Comparing to related prior work, we empirically show that our Markovian
state-space assumption enables faithful and much improved long-term prediction
well beyond the training horizon. Further, our inference model correctly
decomposes frames into objects, even in the presence of occlusions. Tracking
performance is increased significantly over prior art.
| stat |
Weather-inspired ensemble-based probabilistic prediction of COVID-19 | The objective of this work is to predict the spread of COVID-19 starting from
observed data, using a forecast method inspired by probabilistic weather
prediction systems operational today.
Results show that this method works well for China: on day 25 we could have
predicted well the outcome for the next 35 days. The same method has been
applied to Italy and South Korea, and forecasts for the forthcoming weeks are
included in this work. For Italy, forecasts based on data collected up to today
(24 March) indicate that number of observed cases could grow from the current
value of 69,176, to between 101k-180k, with a 50% probability of being between
110k-135k. For South Korea, it suggests that the number of observed cases could
grow from the current value of 9,018 (as of the 23rd of March), to values
between 8,500 and 9,300, with a 50% probability of being between 8,700 and
8,900.
We conclude by suggesting that probabilistic disease prediction systems are
possible and could be developed following key ideas and methods from weather
forecasting. Having access to skilful daily updated forecasts could help taking
better informed decisions on how to manage the spread of diseases such as
COVID-19.
| stat |
Hausdorff Dimension, Heavy Tails, and Generalization in Neural Networks | Despite its success in a wide range of applications, characterizing the
generalization properties of stochastic gradient descent (SGD) in non-convex
deep learning problems is still an important challenge. While modeling the
trajectories of SGD via stochastic differential equations (SDE) under
heavy-tailed gradient noise has recently shed light over several peculiar
characteristics of SGD, a rigorous treatment of the generalization properties
of such SDEs in a learning theoretical framework is still missing. Aiming to
bridge this gap, in this paper, we prove generalization bounds for SGD under
the assumption that its trajectories can be well-approximated by a \emph{Feller
process}, which defines a rich class of Markov processes that include several
recent SDE representations (both Brownian or heavy-tailed) as its special case.
We show that the generalization error can be controlled by the \emph{Hausdorff
dimension} of the trajectories, which is intimately linked to the tail behavior
of the driving process. Our results imply that heavier-tailed processes should
achieve better generalization; hence, the tail-index of the process can be used
as a notion of "capacity metric". We support our theory with experiments on
deep neural networks illustrating that the proposed capacity metric accurately
estimates the generalization error, and it does not necessarily grow with the
number of parameters unlike the existing capacity metrics in the literature.
| stat |
Estimating functional parameters for understanding the impact of weather
and government interventions on COVID-19 outbreak | As the coronavirus disease 2019 (COVID-19) has shown profound effects on
public health and the economy worldwide, it becomes crucial to assess the
impact on the virus transmission and develop effective strategies to address
the challenge. A new statistical model derived from the SIR epidemic model with
functional parameters is proposed to understand the impact of weather and
government interventions on the virus spread and also provide the forecasts of
COVID-19 infections among eight metropolitan areas in the United States. The
model uses Bayesian inference with Gaussian process priors to study the
functional parameters nonparametrically, and sensitivity analysis is adopted to
investigate the main and interaction effects of these factors. This analysis
reveals several important results including the potential interaction effects
between weather and government interventions, which shed new light on the
effective strategies for policymakers to mitigate the COVID-19 outbreak.
| stat |
Semiparametric estimation of structural failure time model in
continuous-time processes | Structural failure time models are causal models for estimating the effect of
time-varying treatments on a survival outcome. G-estimation and artificial
censoring have been proposed to estimate the model parameters in the presence
of time-dependent confounding and administrative censoring. However, most of
existing methods require manually preprocessing data into regularly spaced
data, which may invalidate the subsequent causal analysis. Moreover, the
computation and inference are challenging due to the non-smoothness of
artificial censoring. We propose a class of continuous-time structural failure
time models, which respects the continuous time nature of the underlying data
processes. Under a martingale condition of no unmeasured confounding, we show
that the model parameters are identifiable from potentially infinite estimating
equations. Using the semiparametric efficiency theory, we derive the first
semiparametric doubly robust estimators, in the sense that the estimators are
consistent if either the treatment process model or the failure time model is
correctly specified, but not necessarily both. Moreover, we propose using
inverse probability of censoring weighting to deal with dependent censoring. In
contrast to artificial censoring, our weighting strategy does not introduce
non-smoothness in estimation and ensures that the resampling methods can be
used to make inference.
| stat |
Sparsely Observed Functional Time Series: Estimation and Prediction | Functional time series analysis, whether based on time of frequency domain
methodology, has traditionally been carried out under the assumption of
complete observation of the constituent series of curves, assumed stationary.
Nevertheless, as is often the case with independent functional data, it may
well happen that the data available to the analyst are not the actual sequence
of curves, but relatively few and noisy measurements per curve, potentially at
different locations in each curve's domain. Under this sparse sampling regime,
neither the established estimators of the time series' dynamics, nor their
corresponding theoretical analysis will apply. The subject of this paper is to
tackle the problem of estimating the dynamics and of recovering the latent
process of smooth curves in the sparse regime. Assuming smoothness of the
latent curves, we construct a consistent nonparametric estimator of the series'
spectral density operator and use it develop a frequency-domain recovery
approach, that predicts the latent curve at a given time by borrowing strength
from the (estimated) dynamic correlations in the series across time. Further to
predicting the latent curves from their noisy point samples, the method fills
in gaps in the sequence (curves nowhere sampled), denoises the data, and serves
as a basis for forecasting. Means of providing corresponding confidence bands
are also investigated. A simulation study interestingly suggests that sparse
observation for a longer time period, may be provide better performance than
dense observation for a shorter period, in the presence of smoothness. The
methodology is further illustrated by application to an environmental data set
on fair-weather atmospheric electricity, which naturally leads to a sparse
functional time-series.
| stat |
Smoothed Nested Testing on Directed Acyclic Graphs | We consider the problem of multiple hypothesis testing when there is a
logical nested structure to the hypotheses. When one hypothesis is nested
inside another, the outer hypothesis must be false if the inner hypothesis is
false. We model the nested structure as a directed acyclic graph, including
chain and tree graphs as special cases. Each node in the graph is a hypothesis
and rejecting a node requires also rejecting all of its ancestors. We propose a
general framework for adjusting node-level test statistics using the known
logical constraints. Within this framework, we study a smoothing procedure that
combines each node with all of its descendants to form a more powerful
statistic. We prove a broad class of smoothing strategies can be used with
existing selection procedures to control the familywise error rate, false
discovery exceedance rate, or false discovery rate, so long as the original
test statistics are independent under the null. When the null statistics are
not independent but are derived from positively-correlated normal observations,
we prove control for all three error rates when the smoothing method is
arithmetic averaging of the observations. Simulations and an application to a
real biology dataset demonstrate that smoothing leads to substantial power
gains.
| stat |
A New Spatial Count Data Model with Bayesian Additive Regression Trees
for Accident Hot Spot Identification | The identification of accident hot spots is a central task of road safety
management. Bayesian count data models have emerged as the workhorse method for
producing probabilistic rankings of hazardous sites in road networks.
Typically, these methods assume simple linear link function specifications,
which, however, limit the predictive power of a model. Furthermore, extensive
specification searches are precluded by complex model structures arising from
the need to account for unobserved heterogeneity and spatial correlations.
Modern machine learning (ML) methods offer ways to automate the specification
of the link function. However, these methods do not capture estimation
uncertainty, and it is also difficult to incorporate spatial correlations. In
light of these gaps in the literature, this paper proposes a new spatial
negative binomial model, which uses Bayesian additive regression trees to
endogenously select the specification of the link function. Posterior inference
in the proposed model is made feasible with the help of the Polya-Gamma data
augmentation technique. We test the performance of this new model on a crash
count data set from a metropolitan highway network. The empirical results show
that the proposed model performs at least as well as a baseline spatial count
data model with random parameters in terms of goodness of fit and site ranking
ability.
| stat |
Langevin Monte Carlo without smoothness | Langevin Monte Carlo (LMC) is an iterative algorithm used to generate samples
from a distribution that is known only up to a normalizing constant. The
nonasymptotic dependence of its mixing time on the dimension and target
accuracy is understood mainly in the setting of smooth (gradient-Lipschitz)
log-densities, a serious limitation for applications in machine learning. In
this paper, we remove this limitation, providing polynomial-time convergence
guarantees for a variant of LMC in the setting of nonsmooth log-concave
distributions. At a high level, our results follow by leveraging the implicit
smoothing of the log-density that comes from a small Gaussian perturbation that
we add to the iterates of the algorithm and controlling the bias and variance
that are induced by this perturbation.
| stat |
Spherical Motion Dynamics: Learning Dynamics of Neural Network with
Normalization, Weight Decay, and SGD | In this work, we comprehensively reveal the learning dynamics of neural
network with normalization, weight decay (WD), and SGD (with momentum), named
as Spherical Motion Dynamics (SMD). Most related works study SMD by focusing on
"effective learning rate" in "equilibrium" condition, where weight norm remains
unchanged. However, their discussions on why equilibrium condition can be
reached in SMD is either absent or less convincing. Our work investigates SMD
by directly exploring the cause of equilibrium condition. Specifically, 1) we
introduce the assumptions that can lead to equilibrium condition in SMD, and
prove that weight norm can converge at linear rate with given assumptions; 2)
we propose "angular update" as a substitute for effective learning rate to
measure the evolving of neural network in SMD, and prove angular update can
also converge to its theoretical value at linear rate; 3) we verify our
assumptions and theoretical results on various computer vision tasks including
ImageNet and MSCOCO with standard settings. Experiment results show our
theoretical findings agree well with empirical observations.
| stat |
Robust multivariate methods in Chemometrics | This chapter presents an introduction to robust statistics with applications
of a chemometric nature. Following a description of the basic ideas and
concepts behind robust statistics, including how robust estimators can be
conceived, the chapter builds up to the construction (and use) of robust
alternatives for some methods for multivariate analysis frequently used in
chemometrics, such as principal component analysis and partial least squares.
The chapter then provides an insight into how these robust methods can be used
or extended to classification. To conclude, the issue of validation of the
results is being addressed: it is shown how uncertainty statements associated
with robust estimates, can be obtained.
| stat |
Space-time calibration of wind speed forecasts from regional climate
models | Numerical weather predictions (NWP) are systematically subject to errors due
to the deterministic solutions used by numerical models to simulate the
atmosphere. Statistical postprocessing techniques are widely used nowadays for
NWP calibration. However, time-varying bias is usually not accommodated by such
models. Its calibration performance is also sensitive to the temporal window
used for training. This paper proposes space-time models that extend the main
statistical postprocessing approaches to calibrate NWP model outputs.
Trans-Gaussian random fields are considered to account for meteorological
variables with asymmetric behavior. Data augmentation is used to account for
censuring in the response variable. The benefits of the proposed extensions are
illustrated through the calibration of hourly 10 m wind speed forecasts in
Southeastern Brazil coming from the Eta model.
| stat |
An efficient Bayesian experimental calibration of dynamic thermal models | Experimental calibration of dynamic thermal models is required for model
predictive control and characterization of building energy performance. In
these applications, the uncertainty assessment of the parameter estimates is
decisive; this is why a Bayesian calibration procedure (selection, calibration
and validation) is presented. The calibration is based on an improved
Metropolis-Hastings algorithm suitable for linear and Gaussian state-space
models. The procedure, illustrated on a real house experiment, shows that the
algorithm is more robust to initial conditions than a maximum likelihood
optimization with a quasi-Newton algorithm. Furthermore, when the data are not
informative enough, the use of prior distributions helps to regularize the
problem.
| stat |
Regression Modeling for Recurrent Events Using R Package reReg | Recurrent event analyses have found a wide range of applications in
biomedicine, public health, and engineering, among others, where study subjects
may experience a sequence of event of interest during follow-up. The R package
reReg (Chiou and Huang 2021) offers a comprehensive collection of practical and
easy-to-use tools for regression analysis of recurrent events, possibly with
the presence of an informative terminal event. The regression framework is a
general scale-change model which encompasses the popular Cox-type model, the
accelerated rate model, and the accelerated mean model as special cases.
Informative censoring is accommodated through a subject-specific frailty
without no need for parametric specification. Different regression models are
allowed for the recurrent event process and the terminal event. Also included
are visualization and simulation tools.
| stat |
Improving Interpretable Piecewise Linear Models through Hierarchical
Spatial and Functional Smoothing | Scientists often use simple models with solutions that provide insight into
physical and environmental systems. Although easy to understand, these simple
models often lack the flexibility to accurately model associated data. We focus
on the setting where we have functional data distributed over space, where (1)
the interpretable model can be expressed as a linear combination of basis
functions, (2) the properties of the data-generating process vary spatially,
and (3) the interpretable model is too simple to effectively match the data. In
this manuscript, through our motivating example of modeling snow density, we
develop a framework for functionally smoothing generalized piecewise linear
models while preserving inference on the simple model by projecting a smooth
function into the orthogonal column space of the piecewise linear model.
Moreover, we allow the parameters of the simple model and the functional
smoothing to vary spatially.
We use a snow density model for ice sheets as the motivating application of
this model. The underlying piecewise linear differential equation solution
fails to match several data features. We address these issues with a novel,
physically-constrained regression model for snow density as a function of
depth. The proposed spatially and functionally smoothed snow density model
better fits the data while preserving inference on physical parameters. Lastly,
we use a unique hierarchical, heteroscedastic error model that accounts for
differences between data sources. Using this model, we find significant spatial
variation in the parameters that govern snow densification.
| stat |
Practical Deep Learning with Bayesian Principles | Bayesian methods promise to fix many shortcomings of deep learning, but they
are impractical and rarely match the performance of standard methods, let alone
improve them. In this paper, we demonstrate practical training of deep networks
with natural-gradient variational inference. By applying techniques such as
batch normalisation, data augmentation, and distributed training, we achieve
similar performance in about the same number of epochs as the Adam optimiser,
even on large datasets such as ImageNet. Importantly, the benefits of Bayesian
principles are preserved: predictive probabilities are well-calibrated,
uncertainties on out-of-distribution data are improved, and continual-learning
performance is boosted. This work enables practical deep learning while
preserving benefits of Bayesian principles. A PyTorch implementation is
available as a plug-and-play optimiser.
| stat |
Optimization for L1-Norm Error Fitting via Data Aggregation | We propose a data aggregation-based algorithm with monotonic convergence to a
global optimum for a generalized version of the L1-norm error fitting model
with an assumption of the fitting function. The proposed algorithm generalizes
the recent algorithm in the literature, aggregate and iterative disaggregate
(AID), which selectively solves three specific L1-norm error fitting problems.
With the proposed algorithm, any L1-norm error fitting model can be solved
optimally if it follows the form of the L1-norm error fitting problem and if
the fitting function satisfies the assumption. The proposed algorithm can also
solve multi-dimensional fitting problems with arbitrary constraints on the
fitting coefficients matrix. The generalized problem includes popular models
such as regression and the orthogonal Procrustes problem. The results of the
computational experiment show that the proposed algorithms are faster than the
state-of-the-art benchmarks for L1-norm regression subset selection and L1-norm
regression over a sphere. Further, the relative performance of the proposed
algorithm improves as data size increases.
| stat |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.