title
stringlengths 6
244
| abstract
stringlengths 19
5.09k
| label
stringclasses 10
values |
|---|---|---|
Exponential Weights on the Hypercube in Polynomial Time | We study a general online linear optimization problem(OLO). At each round, a
subset of objects from a fixed universe of $n$ objects is chosen, and a linear
cost associated with the chosen subset is incurred. To measure the performance
of our algorithms, we use the notion of regret which is the difference between
the total cost incurred over all iterations and the cost of the best fixed
subset in hindsight. We consider Full Information and Bandit feedback for this
problem. This problem is equivalent to OLO on the $\{0,1\}^n$ hypercube. The
Exp2 algorithm and its bandit variant are commonly used strategies for this
problem. It was previously unknown if it is possible to run Exp2 on the
hypercube in polynomial time.
In this paper, we present a polynomial time algorithm called PolyExp for OLO
on the hypercube. We show that our algorithm is equivalent Exp2 on $\{0,1\}^n$,
Online Mirror Descent(OMD), Follow The Regularized Leader(FTRL) and Follow The
Perturbed Leader(FTPL) algorithms. We show PolyExp achieves expected regret
bound that is a factor of $\sqrt{n}$ better than Exp2 in the full information
setting under $L_\infty$ adversarial losses. Because of the equivalence of
these algorithms, this implies an improvement on Exp2's regret bound in full
information. We also show matching regret lower bounds. Finally, we show how to
use PolyExp on the $\{-1,+1\}^n$ hypercube, solving an open problem in Bubeck
et al (COLT 2012).
| stat |
A parameter-free population-dynamical approach to health workforce
supply forecasting of EU countries | Many countries face challenges like impending retirement waves, negative
population growth, or a suboptimal distribution of resources across medical
sectors and fields in supplying their healthcare systems with adequate
staffing. An increasing number of countries therefore employs quantitative
approaches in health workforce supply forecasting. However, these models are
often of limited usability as they either require extensive individual-level
data or become too simplistic to capture key demographic or epidemiological
factors. We propose a novel population-dynamical and stock-flow-consistent
approach to health workforce supply forecasting complex enough to address
dynamically changing behaviors while requiring only publicly available
timeseries data for complete calibration. We apply the model to 21 European
countries to forecast the supply of generalist and specialist physicians until
2040. Compared to staffing levels required to keep the physician density
constant at 2016 levels, in many countries we find a significant trend toward
decreasing density for generalist physicians at the expense of increasing
densities for specialists. These trends are exacerbated in many Southern and
Eastern European countries by expectations of negative population growth. For
the example of Austria we generalize our approach to a multi-professional,
multi-regional and multi-sectoral model and find a suboptimal distribution in
the supply of contracted versus non-contracted physicians. It is of the utmost
importance to devise tools for decision makers to influence the allocation and
supply of physicians across fields and sectors to combat imbalances.
| stat |
Learning disentangled representations with the Wasserstein Autoencoder | Disentangled representation learning has undoubtedly benefited from objective
function surgery. However, a delicate balancing act of tuning is still required
in order to trade off reconstruction fidelity versus disentanglement. Building
on previous successes of penalizing the total correlation in the latent
variables, we propose TCWAE (Total Correlation Wasserstein Autoencoder).
Working in the WAE paradigm naturally enables the separation of the
total-correlation term, thus providing disentanglement control over the learned
representation, while offering more flexibility in the choice of reconstruction
cost. We propose two variants using different KL estimators and perform
extensive quantitative comparisons on data sets with known generative factors,
showing competitive results relative to state-of-the-art techniques. We further
study the trade off between disentanglement and reconstruction on
more-difficult data sets with unknown generative factors, where the flexibility
of the WAE paradigm in the reconstruction term improves reconstructions.
| stat |
Continuous-time multi-state capture-recapture models | Multi-state capture-recapture data comprise individual-specific sighting
histories together with information on individuals' states related, for
example, to breeding status, infection level, or geographical location. Such
data are often analysed using the Arnason-Schwarz model, where transitions
between states are modelled using a discrete-time Markov chain, making the
model most easily applicable to regular time series. When time intervals
between capture occasions are not of equal length, more complex time-dependent
constructions may be required, increasing the number of parameters to estimate,
decreasing interpretability, and potentially leading to reduced precision. Here
we develop a novel continuous-time multi-state model that can be regarded as an
analogue of the Arnason-Schwarz model for irregularly sampled data. Statistical
inference is carried out by regarding the capture-recapture data as
realisations from a continuous-time hidden Markov model, which allows the
associated efficient algorithms to be used for maximum likelihood estimation
and state decoding. To illustrate the feasibility of the modelling framework,
we use a long-term survey of bottlenose dolphins where capture occasion are not
regularly spaced through time. Here we are particularly interested in seasonal
effects on the movement rates of the dolphins along the Scottish east coast.
The results reveal seasonal movement patterns between two core areas of their
range, providing information that will inform conservation management.
| stat |
Product Partition Dynamic Generalized Linear Models | Detection and modeling of change-points in time-series can be considerably
challenging. In this paper we approach this problem by incorporating the class
of Dynamic Generalized Linear Models (DGLM) into the well know class of Product
Partition Models (PPM). This new methodology, that we call DGLM-PPM, extends
the PPM to distributions within the Exponential Family while also retaining the
flexibility of the DGLM class. It also provides a framework for Bayesian
multiple change-point detection in dynamic regression models. Inference on the
DGLM-PPM follow the steps of evolution and updating of the DGLM class. A Gibbs
Sampler scheme with an Adaptive Rejection Metropolis Sampling (ARMS) step
appended is used to compute posterior estimates of the relevant quantities. A
simulation study shows that the proposed model provides reasonable estimates of
the dynamic parameters and also assigns high change-point probabilities to the
breaks introduced in the artificial data generated for this work. We also
present a real life data example that highlights the superiority of the
DGLM-PPM over the conventional DGLM in both in-sample and out-of-sample
goodness of fit measures.
| stat |
Integrated Continuous-time Hidden Markov Models | Motivated by applications in movement ecology, in this paper I propose a new
class of integrated continuous-time hidden Markov models in which each
observation depends on the underlying state of the process over the whole
interval since the previous observation, not only on its current state. This
class gives a new representation of a range of existing models, including some
widely applied switching diffusion models. I show that under appropriate
conditioning, a model in this class can be regarded as a conventional hidden
Markov model, enabling use of the Forward Algorithm for efficient evaluation of
its likelihood without sampling of its state sequence. This leads to an
algorithm for inference which is more efficient, and scales better with the
amount of data, than existing methods. This is demonstrated and quantified in
some applications to animal movement data and some related simulation
experiments.
| stat |
Identifying Gene-environment interactions with robust marginal Bayesian
variable selection | In high-throughput genetics studies, an important aim is to identify
gene-environment interactions associated with the clinical outcomes. Recently,
multiple marginal penalization methods have been developed and shown to be
effective in G$\times$E studies. However, within the Bayesian framework,
marginal variable selection has not received much attention. In this study, we
propose a novel marginal Bayesian variable selection method for G$\times$E
studies. In particular, our marginal Bayesian method is robust to data
contamination and outliers in the outcome variables. With the incorporation of
spike-and-slab priors, we have implemented the Gibbs sampler based on MCMC. The
proposed method outperforms a number of alternatives in extensive simulation
studies. The utility of the marginal robust Bayesian variable selection method
has been further demonstrated in the case studies using data from the Nurse
Health Study (NHS). Some of the identified main and interaction effects from
the real data analysis have important biological implications.
| stat |
OpenML Benchmarking Suites | Machine learning research depends on objectively interpretable, comparable,
and reproducible algorithm benchmarks. Therefore, we advocate the use of
curated, comprehensive suites of machine learning tasks to standardize the
setup, execution, and reporting of benchmarks. We enable this through software
tools that help to create and leverage these benchmarking suites. These are
seamlessly integrated into the OpenML platform, and accessible through
interfaces in Python, Java, and R. OpenML benchmarking suites are (a) easy to
use through standardized data formats, APIs, and client libraries; (b)
machine-readable, with extensive meta-information on the included datasets; and
(c) allow benchmarks to be shared and reused in future studies. We also present
a first, carefully curated and practical benchmarking suite for classification:
the OpenML Curated Classification benchmarking suite 2018 (OpenML-CC18).
| stat |
Noisy Adaptive Group Testing using Bayesian Sequential Experimental
Design | When the infection prevalence of a disease is low, Dorfman showed 80 years
ago that testing groups of people can prove more efficient than testing people
individually. Our goal in this paper is to propose new group testing algorithms
that can operate in a noisy setting (tests can be mistaken) to decide
adaptively (looking at past results) which groups to test next, with the goal
to converge to a good detection, as quickly, and with as few tests as possible.
We cast this problem as a Bayesian sequential experimental design problem.
Using the posterior distribution of infection status vectors for $n$ patients,
given observed tests carried out so far, we seek to form groups that have a
maximal utility. We consider utilities such as mutual information, but also
quantities that have a more direct relevance to testing, such as the AUC of the
ROC curve of the test. Practically, the posterior distributions on $\{0,1\}^n$
are approximated by sequential Monte Carlo (SMC) samplers and the utility
maximized by a greedy optimizer. Our procedures show in simulations significant
improvements over both adaptive and non-adaptive baselines, and are far more
efficient than individual tests when disease prevalence is low. Additionally,
we show empirically that loopy belief propagation (LBP), widely regarded as the
SoTA decoder to decide whether an individual is infected or not given previous
tests, can be unreliable and exhibit oscillatory behavior. Our SMC decoder is
more reliable, and can improve the performance of other group testing
algorithms.
| stat |
The Randomized Elliptical Potential Lemma with an Application to Linear
Thompson Sampling | In this note, we introduce a randomized version of the well-known elliptical
potential lemma that is widely used in the analysis of algorithms in sequential
learning and decision-making problems such as stochastic linear bandits. Our
randomized elliptical potential lemma relaxes the Gaussian assumption on the
observation noise and on the prior distribution of the problem parameters. We
then use this generalization to prove an improved Bayesian regret bound for
Thompson sampling for the linear stochastic bandits with changing action sets
where prior and noise distributions are general. This bound is minimax optimal
up to constants.
| stat |
From scenario-based seismic hazard to scenario-based landslide hazard:
fast-forwarding to the future via statistical simulations | Ground motion scenarios exists for most of the seismically active areas
around the globe. They essentially correspond to shaking level maps at given
earthquake return times which are used as reference for the likely areas under
threat from future ground displacements. Being landslides in seismically
actively regions closely controlled by the ground motion, one would expect that
landslide susceptibility maps should change as the ground motion patterns
change in space and time. However, so far, statistically-based landslide
susceptibility assessments have primarily been used as time-invariant.In other
words, the vast majority of the statistical models does not include the
temporal effect of the main trigger in future landslide scenarios. In this
work, we present an approach aimed at filling this gap, bridging current
practices in the seismological community to those in the geomorphological and
statistical ones. More specifically, we select an earthquake-induced landslide
inventory corresponding to the 1994 Northridge earthquake and build a Bayesian
Generalized Additive Model of the binomial family, featuring common
morphometric and thematic covariates as well as the Peak Ground Acceleration
generated by the Northridge earthquake. Once each model component has been
estimated, we have run 1000 simulations for each of the 217 possible ground
motion scenarios for the study area.
From each batch of 1000 simulations, we have estimated the mean and 95\%
Credible Interval to represent the mean susceptibility pattern under a specific
earthquake scenario, together with its uncertainty level. Because each
earthquake scenario has a specific return time, our simulations allow to
incorporate the temporal dimension into any susceptibility model, therefore
driving the results toward the definition of landslide hazard.
| stat |
Power-Expected-Posterior Priors as Mixtures of g-Priors | One of the main approaches used to construct prior distributions for
objective Bayes methods is the concept of random imaginary observations. Under
this setup, the expected-posterior prior (EPP) offers several advantages, among
which it has a nice and simple interpretation and provides an effective way to
establish compatibility of priors among models. In this paper, we study the
power-expected posterior prior as a generalization to the EPP in objective
Bayesian model selection under normal linear models. We prove that it can be
represented as a mixture of $g$-prior, like a wide range of prior distributions
under normal linear models, and thus posterior distributions and Bayes factors
are derived in closed form, keeping therefore computational tractability.
Comparisons with other mixtures of $g$-prior are made and emphasis is given in
the posterior distribution of g and its effect on Bayesian model selection and
model averaging.
| stat |
Uncertainty representation for early phase clinical test evaluations: a
case study | In early clinical test evaluations the potential benefits of the introduction
of a new technology into the healthcare system are assessed in the challenging
situation of limited available empirical data. The aim of these evaluations is
to provide additional evidence for the decision maker, who is typically a
funder or the company developing the test, to evaluate which technologies
should progress to the next stage of evaluation. In this paper we consider the
evaluation of a diagnostic test for patients suffering from Chronic Obstructive
Pulmonary Disease (COPD). We describe the use of graphical models, prior
elicitation and uncertainty analysis to provide the required evidence to allow
the test to progress to the next stage of evaluation. We specifically discuss
inferring an influence diagram from a care pathway and conducting an
elicitation exercise to allow specification of prior distributions over all
model parameters. We describe the uncertainty analysis, via Monte Carlo
simulation, which allowed us to demonstrate that the potential value of the
test was robust to uncertainties. This paper provides a case study illustrating
how a careful Bayesian analysis can be used to enhance early clinical test
evaluations.
| stat |
Complexity as Causal Information Integration | Complexity measures in the context of the Integrated Information Theory of
consciousness try to quantify the strength of the causal connections between
different neurons. This is done by minimizing the KL-divergence between a full
system and one without causal connections. Various measures have been proposed
and compared in this setting. We will discuss a class of information geometric
measures that aim at assessing the intrinsic causal influences in a system. One
promising candidate of these measures, denoted by $\Phi_{CIS}$, is based on
conditional independence statements and does satisfy all of the properties that
have been postulated as desirable. Unfortunately it does not have a graphical
representation which makes it less intuitive and difficult to analyze. We
propose an alternative approach using a latent variable which models a common
exterior influence. This leads to a measure $\Phi_{CII}$, Causal Information
Integration, that satisfies all of the required conditions. Our measure can be
calculated using an iterative information geometric algorithm, the
em-algorithm. Therefore we are able to compare its behavior to existing
integrated information measures.
| stat |
Considerations for developing predictive models of crime and new methods
for measuring their accuracy | Developing spatio-temporal crime prediction models, and to a lesser extent,
developing measures of accuracy and operational efficiency for them, has been
an active area of research for almost two decades. Despite calls for rigorous
and independent evaluations of model performance, such studies have been few
and far between. In this paper, we argue that studies should focus not on
finding the one predictive model or the one measure that is the most
appropriate at all times, but instead on careful consideration of several
factors that affect the choice of the model and the choice of the measure, to
find the best measure and the best model for the problem at hand. We argue that
because each problem is unique, it is important to develop measures that
empower the practitioner with the ability to input the choices and preferences
that are most appropriate for the problem at hand. We develop a new measure
called the penalized predictive accuracy index (PPAI) which imparts such
flexibility. We also propose the use of the expected utility function to
combine multiple measures in a way that is appropriate for a given problem in
order to assess the models against multiple criteria. We further propose the
use of the average logarithmic score (ALS) measure that is appropriate for many
crime models and measures accuracy differently than existing measures. These
measures can be used alongside existing measures to provide a more
comprehensive means of assessing the accuracy and potential utility of
spatio-temporal crime prediction models.
| stat |
Bayesian Functional Principal Components Analysis via Variational
Message Passing | Functional principal components analysis is a popular tool for inference on
functional data. Standard approaches rely on an eigendecomposition of a
smoothed covariance surface in order to extract the orthonormal functions
representing the major modes of variation. This approach can be a
computationally intensive procedure, especially in the presence of large
datasets with irregular observations. In this article, we develop a Bayesian
approach, which aims to determine the Karhunen-Lo\`eve decomposition directly
without the need to smooth and estimate a covariance surface. More
specifically, we develop a variational Bayesian algorithm via message passing
over a factor graph, which is more commonly referred to as variational message
passing. Message passing algorithms are a powerful tool for compartmentalizing
the algebra and coding required for inference in hierarchical statistical
models. Recently, there has been much focus on formulating variational
inference algorithms in the message passing framework because it removes the
need for rederiving approximate posterior density functions if there is a
change to the model. Instead, model changes are handled by changing specific
computational units, known as fragments, within the factor graph. We extend the
notion of variational message passing to functional principal components
analysis. Indeed, this is the first article to address a functional data model
via variational message passing. Our approach introduces two new fragments that
are necessary for Bayesian functional principal components analysis. We present
the computational details, a set of simulations for assessing accuracy and
speed and an application to United States temperature data.
| stat |
Analyzing Brain Circuits in Population Neuroscience: A Case to Be a
Bayesian | Functional connectivity fingerprints are among today's best choices to obtain
a faithful sampling of an individual's brain and cognition in health and
disease. Here we make a case for key advantages of analyzing such connectome
profiles using Bayesian analysis strategies. They (i) afford full probability
estimates of the studied neurocognitive phenomenon (ii) provide analytical
machinery to separate methodological uncertainty and biological variability in
a coherent manner (iii) usher towards avenues to go beyond classical
null-hypothesis significance testing and (iv) enable estimates of credibility
around all model parameters at play and thus enable predictions with
uncertainty intervals for single subject. We pick research questions about
autism spectrum disorder as a recurring theme to illustrate our methodological
arguments.
| stat |
A new INARMA(1, 1) model with Poisson marginals | We suggest an INARMA(1, 1) model with Poisson marginals which extends the
INAR(1) in a similar way as the INGARCH(1, 1) does for the INARCH(1) model. The
new model is equivalent to a binomially thinned INAR(1) process. This allows us
to obtain some of its stochastic properties and use inference methods for
hidden Markov models. The model is compared to various other models in two case
studies.
| stat |
An introduction to Bent Jorgensen's ideas | We briefly expose some key aspects of the theory and use of dispersion
models, for which Bent Jorgensen played a crucial role as a driving force and
an inspiration source. Starting with the general notion of dispersion models,
built using minimalistic mathematical assumptions, we specialize in two classes
of families of distributions with different statistical flavors: exponential
dispersion and proper dispersion models. The construction of dispersion models
involves the solution of integral equations that are, in general, untractable.
These difficulties disappear when a more mathematical structure is assumed: it
reduces to the calculation of a moment generating function or of a
Riemann-Stieltjes integral for the exponential dispersion and the proper
dispersion models, respectively. A new technique for constructing dispersion
models based on characteristic functions is introduced turning the integral
equations above into a tractable convolution equation and yielding examples of
dispersion models that are neither proper dispersion nor exponential dispersion
models. A corollary is that the cardinality of regular and non-regular
dispersion models are both large.
Some selected applications are discussed including exponential families
non-linear models (for which generalized linear models are particular cases)
and several models for clustered and dependent data based on a latent Levy
process.
| stat |
Separable Layers Enable Structured Efficient Linear Substitutions | In response to the development of recent efficient dense layers, this paper
shows that something as simple as replacing linear components in pointwise
convolutions with structured linear decompositions also produces substantial
gains in the efficiency/accuracy tradeoff. Pointwise convolutions are fully
connected layers and are thus prepared for replacement by structured
transforms. Networks using such layers are able to learn the same tasks as
those using standard convolutions, and provide Pareto-optimal benefits in
efficiency/accuracy, both in terms of computation (mult-adds) and parameter
count (and hence memory). Code is available at
https://github.com/BayesWatch/deficient-efficient.
| stat |
Random Partition Models for Microclustering Tasks | Traditional Bayesian random partition models assume that the size of each
cluster grows linearly with the number of data points. While this is appealing
for some applications, this assumption is not appropriate for other tasks such
as entity resolution, modeling of sparse networks, and DNA sequencing tasks.
Such applications require models that yield clusters whose sizes grow
sublinearly with the total number of data points -- the microclustering
property. Motivated by these issues, we propose a general class of random
partition models that satisfy the microclustering property with
well-characterized theoretical properties. Our proposed models overcome major
limitations in the existing literature on microclustering models, namely a lack
of interpretability, identifiability, and full characterization of model
asymptotic properties. Crucially, we drop the classical assumption of having an
exchangeable sequence of data points, and instead assume an exchangeable
sequence of clusters. In addition, our framework provides flexibility in terms
of the prior distribution of cluster sizes, computational tractability, and
applicability to a large number of microclustering tasks. We establish
theoretical properties of the resulting class of priors, where we characterize
the asymptotic behavior of the number of clusters and of the proportion of
clusters of a given size. Our framework allows a simple and efficient Markov
chain Monte Carlo algorithm to perform statistical inference. We illustrate our
proposed methodology on the microclustering task of entity resolution, where we
provide a simulation study and real experiments on survey panel data.
| stat |
The Online Saddle Point Problem and Online Convex Optimization with
Knapsacks | We study the online saddle point problem, an online learning problem where at
each iteration a pair of actions need to be chosen without knowledge of the
current and future (convex-concave) payoff functions. The objective is to
minimize the gap between the cumulative payoffs and the saddle point value of
the aggregate payoff function, which we measure using a metric called
"SP-Regret". The problem generalizes the online convex optimization framework
but here we must ensure both players incur cumulative payoffs close to that of
the Nash equilibrium of the sum of the games. We propose an algorithm that
achieves SP-Regret proportional to $\sqrt{\ln(T)T}$ in the general case, and
$\log(T)$ SP-Regret for the strongly convex-concave case. We also consider the
special case where the payoff functions are bilinear and the decision sets are
the probability simplex. In this setting we are able to design algorithms that
reduce the bounds on SP-Regret from a linear dependence in the dimension of the
problem to a \textit{logarithmic} one. We also study the problem under bandit
feedback and provide an algorithm that achieves sublinear SP-Regret. We then
consider an online convex optimization with knapsacks problem motivated by a
wide variety of applications such as: dynamic pricing, auctions, and
crowdsourcing. We relate this problem to the online saddle point problem and
establish $O(\sqrt{T})$ regret using a primal-dual algorithm.
| stat |
Model selection for count timeseries with applications in forecasting
number of trips in bike-sharing systems and its volatility | Forecasting the number of trips in bike-sharing systems and its volatility
over time is crucial for planning and optimizing such systems. This paper
develops timeseries models to forecast hourly count timeseries data, and
estimate its volatility. Such models need to take into account the complex
patterns over various temporal scales including hourly, daily, weekly and
annual as well as the temporal correlation. To capture this complex structure,
a large number of parameters are needed. Here a structural model selection
approach is utilized to choose the parameters. This method explores the
parameter space for a group of covariates at each step. These groups of
covariate are constructed to represent a particular structure in the model. The
statistical models utilized are extensions of Generalized Linear Models to
timeseries data. One challenge in using such models is the explosive behavior
of the simulated values. To address this issue, we develop a technique which
relies on damping the simulated value, if it falls outside of an admissible
interval. The admissible interval is defined using measures of variability of
the left and right tails. A new definition of outliers is proposed based on
these variability measures. This new definition is shown to be useful in the
context of asymmetric distributions.
| stat |
Classification using Ensemble Learning under Weighted Misclassification
Loss | Binary classification rules based on covariates typically depend on simple
loss functions such as zero-one misclassification. Some cases may require more
complex loss functions. For example, individual-level monitoring of
HIV-infected individuals on antiretroviral therapy (ART) requires periodic
assessment of treatment failure, defined as having a viral load (VL) value
above a certain threshold. In some resource limited settings, VL tests may be
limited by cost or technology, and diagnoses are based on other clinical
markers. Depending on scenario, higher premium may be placed on avoiding
false-positives which brings greater cost and reduced treatment options. Here,
the optimal rule is determined by minimizing a weighted misclassification
loss/risk.
We propose a method for finding and cross-validating optimal binary
classification rules under weighted misclassification loss. We focus on rules
comprising a prediction score and an associated threshold, where the score is
derived using an ensemble learner. Simulations and examples show that our
method, which derives the score and threshold jointly, more accurately
estimates overall risk and has better operating characteristics compared with
methods that derive the score first and the cutoff conditionally on the score
especially for finite samples.
| stat |
Does mitigating ML's impact disparity require treatment disparity? | Following related work in law and policy, two notions of disparity have come
to shape the study of fairness in algorithmic decision-making. Algorithms
exhibit treatment disparity if they formally treat members of protected
subgroups differently; algorithms exhibit impact disparity when outcomes differ
across subgroups, even if the correlation arises unintentionally. Naturally, we
can achieve impact parity through purposeful treatment disparity. In one thread
of technical work, papers aim to reconcile the two forms of parity proposing
disparate learning processes (DLPs). Here, the learning algorithm can see group
membership during training but produce a classifier that is group-blind at test
time. In this paper, we show theoretically that: (i) When other features
correlate to group membership, DLPs will (indirectly) implement treatment
disparity, undermining the policy desiderata they are designed to address; (ii)
When group membership is partly revealed by other features, DLPs induce
within-class discrimination; and (iii) In general, DLPs provide a suboptimal
trade-off between accuracy and impact parity. Based on our technical analysis,
we argue that transparent treatment disparity is preferable to occluded methods
for achieving impact parity. Experimental results on several real-world
datasets highlight the practical consequences of applying DLPs vs. per-group
thresholds.
| stat |
Modelling and predicting the spatio-temporal spread of Coronavirus
disease 2019 (COVID-19) in Italy | Official freely available data about the number of infected at the finest
possible level of spatial areal aggregation (Italian provinces) are used to
model the spatio-temporal distribution of COVID-19 infections at local level.
Data time horizon ranges from 26 February 20020, which is the date when the
first case not directly connected with China has been discovered in northern
Italy, to 18 March 2020. An endemic-epidemic multivariate time-series
mixed-effects generalized linear model for areal disease counts has been
implemented to understand and predict spatio-temporal diffusion of the
phenomenon. Previous literature has shown that these class of models provide
reliable predictions of infectious diseases in time and space. Three
subcomponents characterize the estimated model. The first is related to the
evolution of the disease over time; the second is characterized by transmission
of the illness among inhabitants of the same province; the third remarks the
effects of spatial neighbourhood and try to capture the contagion effects of
nearby areas. Focusing on the aggregated time-series of the daily counts in
Italy, the contribution of any of the three subcomponents do not dominate on
the others and our predictions are excellent for the whole country, with an
error of 3 per thousand compared to the late available data. At local level,
instead, interesting distinct patterns emerge. In particular, the provinces
first concerned by containment measures are those that are not affected by the
effects of spatial neighbours. On the other hand, for the provinces the are
currently strongly affected by contagions, the component accounting for the
spatial interaction with surrounding areas is prevalent. Moreover, the proposed
model provides good forecasts of the number of infections at local level while
controlling for delayed reporting.
| stat |
Adaptive Geo-Topological Independence Criterion | Testing two potentially multivariate variables for statistical dependence on
the basis finite samples is a fundamental statistical challenge. Here we
explore a family of tests that adapt to the complexity of the relationship
between the variables, promising robust power across scenarios. Building on the
distance correlation, we introduce a family of adaptive independence criteria
based on nonlinear monotonic transformations of distances. We show that these
criteria, like the distance correlation and RKHS-based criteria, provide
dependence indicators. We propose a class of adaptive (multi-threshold) test
statistics, which form the basis for permutation tests. These tests empirically
outperform some of the established tests in average and worst-case statistical
sensitivity across a range of univariate and multivariate relationships, offer
useful insights to the data and may deserve further exploration.
| stat |
Massive parallelization boosts big Bayesian multidimensional scaling | Big Bayes is the computationally intensive co-application of big data and
large, expressive Bayesian models for the analysis of complex phenomena in
scientific inference and statistical learning. Standing as an example, Bayesian
multidimensional scaling (MDS) can help scientists learn viral trajectories
through space-time, but its computational burden prevents its wider use.
Crucial MDS model calculations scale quadratically in the number of
observations. We partially mitigate this limitation through massive
parallelization using multi-core central processing units, instruction-level
vectorization and graphics processing units (GPUs). Fitting the MDS model using
Hamiltonian Monte Carlo, GPUs can deliver more than 100-fold speedups over
serial calculations and thus extend Bayesian MDS to a big data setting. To
illustrate, we employ Bayesian MDS to infer the rate at which different
seasonal influenza virus subtypes use worldwide air traffic to spread around
the globe. We examine 5392 viral sequences and their associated 14 million
pairwise distances arising from the number of commercial airline seats per year
between viral sampling locations. To adjust for shared evolutionary history of
the viruses, we implement a phylogenetic extension to the MDS model and learn
that subtype H3N2 spreads most effectively, consistent with its epidemic
success relative to other seasonal influenza subtypes. Finally, we provide
MassiveMDS, an open-source, stand-alone C++ library and rudimentary R package,
and discuss program design and high-level implementation with an emphasis on
important aspects of computing architecture that become relevant at scale.
| stat |
A Framework for Crop Price Forecasting in Emerging Economies by
Analyzing the Quality of Time-series Data | Accuracy of crop price forecasting techniques is important because it enables
the supply chain planners and government bodies to take appropriate actions by
estimating market factors such as demand and supply. In emerging economies such
as India, the crop prices at marketplaces are manually entered every day, which
can be prone to human-induced errors like the entry of incorrect data or entry
of no data for many days. In addition to such human prone errors, the
fluctuations in the prices itself make the creation of stable and robust
forecasting solution a challenging task. Considering such complexities in crop
price forecasting, in this paper, we present techniques to build robust crop
price prediction models considering various features such as (i) historical
price and market arrival quantity of crops, (ii) historical weather data that
influence crop production and transportation, (iii) data quality-related
features obtained by performing statistical analysis. We additionally propose a
framework for context-based model selection and retraining considering factors
such as model stability, data quality metrics, and trend analysis of crop
prices. To show the efficacy of the proposed approach, we show experimental
results on two crops - Tomato and Maize for 14 marketplaces in India and
demonstrate that the proposed approach not only improves accuracy metrics
significantly when compared against the standard forecasting techniques but
also provides robust models.
| stat |
Modal Principal Component Analysis | Principal component analysis (PCA) is a widely used method for data
processing, such as for dimension reduction and visualization. Standard PCA is
known to be sensitive to outliers, and thus, various robust PCA methods have
been proposed. It has been shown that the robustness of many statistical
methods can be improved using mode estimation instead of mean estimation,
because mode estimation is not significantly affected by the presence of
outliers. Thus, this study proposes a modal principal component analysis
(MPCA), which is a robust PCA method based on mode estimation. The proposed
method finds the minor component by estimating the mode of the projected data
points. As theoretical contribution, probabilistic convergence property,
influence function, finite-sample breakdown point and its lower bound for the
proposed MPCA are derived. The experimental results show that the proposed
method has advantages over the conventional methods.
| stat |
A note on the amount of information borrowed from external data in
hybrid controlled trials with time-to-event outcomes | In situations where it is difficult to enroll patients in randomized
controlled trials, external data can improve efficiency and feasibility. In
such cases, adaptive trial designs could be used to decrease enrollment in the
control arm of the trial by updating the randomization ratio at the interim
analysis. Updating the randomization ratio requires an estimate of the amount
of information effectively borrowed from external data, which is typically done
with a linear approximation. However, this linear approximation is not always a
reliable estimate, which could potentially lead to sub-optimal randomization
ratio updates. In this note, we highlight this issue through simulations for
exponential time-to-event outcomes, because in this simple setting there is an
exact solution available for comparison. We also propose a potential
generalization that could complement the linear approximation in more complex
settings, discuss challenges for this generalization, and recommend best
practices for computing and interpreting estimates of the effective number of
events borrowed.
| stat |
Exploring Spatio-Temporal and Cross-Type Correlations for Crime
Prediction | Crime prediction plays an impactful role in enhancing public security and
sustainable development of urban. With recent advances in data collection and
integration technologies, a large amount of urban data with rich crime-related
information and fine-grained spatio-temporal logs has been recorded. Such
helpful information can boost our understandings about the temporal evolution
and spatial factors of urban crimes and can enhance accurate crime prediction.
In this paper, we perform crime prediction exploiting the cross-type and
spatio-temporal correlations of urban crimes. In particular, we verify the
existence of correlations among different types of crime from temporal and
spatial perspectives, and propose a coherent framework to mathematically model
these correlations for crime prediction. The extensive experimental results on
real-world data validate the effectiveness of the proposed framework. Further
experiments have been conducted to understand the importance of different
correlations in crime prediction.
| stat |
Learning Hierarchical Interactions at Scale: A Convex Optimization
Approach | In many learning settings, it is beneficial to augment the main features with
pairwise interactions. Such interaction models can be often enhanced by
performing variable selection under the so-called strong hierarchy constraint:
an interaction is non-zero only if its associated main features are non-zero.
Existing convex optimization based algorithms face difficulties in handling
problems where the number of main features $p \sim 10^3$ (with total number of
features $\sim p^2$). In this paper, we study a convex relaxation which
enforces strong hierarchy and develop a highly scalable algorithm based on
proximal gradient descent. We introduce novel screening rules that allow for
solving the complicated proximal problem in parallel. In addition, we introduce
a specialized active-set strategy with gradient screening for avoiding costly
gradient computations. The framework can handle problems having dense design
matrices, with $p = 50,000$ ($\sim 10^9$ interactions)---instances that are
much larger than current state of the art. Experiments on real and synthetic
data suggest that our toolkit hierScale outperforms the state of the art in
terms of prediction and variable selection and can achieve over a 4900x
speed-up.
| stat |
Generalisation dynamics of online learning in over-parameterised neural
networks | Deep neural networks achieve stellar generalisation on a variety of problems,
despite often being large enough to easily fit all their training data. Here we
study the generalisation dynamics of two-layer neural networks in a
teacher-student setup, where one network, the student, is trained using
stochastic gradient descent (SGD) on data generated by another network, called
the teacher. We show how for this problem, the dynamics of SGD are captured by
a set of differential equations. In particular, we demonstrate analytically
that the generalisation error of the student increases linearly with the
network size, with other relevant parameters held constant. Our results
indicate that achieving good generalisation in neural networks depends on the
interplay of at least the algorithm, its learning rate, the model architecture,
and the data set.
| stat |
Hierarchical Gaussian Process Priors for Bayesian Neural Network Weights | Probabilistic neural networks are typically modeled with independent weight
priors, which do not capture weight correlations in the prior and do not
provide a parsimonious interface to express properties in function space. A
desirable class of priors would represent weights compactly, capture
correlations between weights, facilitate calibrated reasoning about
uncertainty, and allow inclusion of prior knowledge about the function space
such as periodicity or dependence on contexts such as inputs. To this end, this
paper introduces two innovations: (i) a Gaussian process-based hierarchical
model for network weights based on unit embeddings that can flexibly encode
correlated weight structures, and (ii) input-dependent versions of these weight
priors that can provide convenient ways to regularize the function space
through the use of kernels defined on contextual inputs. We show these models
provide desirable test-time uncertainty estimates on out-of-distribution data,
demonstrate cases of modeling inductive biases for neural networks with kernels
which help both interpolation and extrapolation from training data, and
demonstrate competitive predictive performance on an active learning benchmark.
| stat |
Sparse Bayesian vector autoregressions in huge dimensions | We develop a Bayesian vector autoregressive (VAR) model with multivariate
stochastic volatility that is capable of handling vast dimensional information
sets. Three features are introduced to permit reliable estimation of the model.
First, we assume that the reduced-form errors in the VAR feature a factor
stochastic volatility structure, allowing for conditional equation-by-equation
estimation. Second, we apply recently developed global-local shrinkage priors
to the VAR coefficients to cure the curse of dimensionality. Third, we utilize
recent innovations to efficiently sample from high-dimensional multivariate
Gaussian distributions. This makes simulation-based fully Bayesian inference
feasible when the dimensionality is large but the time series length is
moderate. We demonstrate the merits of our approach in an extensive simulation
study and apply the model to US macroeconomic data to evaluate its forecasting
capabilities.
| stat |
Fast Gaussian Process Based Gradient Matching for Parameter
Identification in Systems of Nonlinear ODEs | Parameter identification and comparison of dynamical systems is a challenging
task in many fields. Bayesian approaches based on Gaussian process regression
over time-series data have been successfully applied to infer the parameters of
a dynamical system without explicitly solving it. While the benefits in
computational cost are well established, a rigorous mathematical framework has
been missing. We offer a novel interpretation which leads to a better
understanding and improvements in state-of-the-art performance in terms of
accuracy for nonlinear dynamical systems.
| stat |
Bayesian nonparametric dynamic hazard rates in evolutionary life tables | In the study of life tables the random variable of interest is usually
assumed discrete since mortality rates are studied for integer ages. In dynamic
life tables a time domain is included to account for the evolution effect of
the hazard rates in time. In this article we follow a survival analysis
approach and use a nonparametric description of the hazard rates. We construct
a discrete time stochastic processes that reflects dependence across age as
well as in time. This process is used as a bayesian nonparametric prior
distribution for the hazard rates for the study of evolutionary life tables.
Prior properties of the process are studied and posterior distributions are
derived. We present a simulation study, with the inclusion of right censored
observations, as well as a real data analysis to show the performance of our
model.
| stat |
Parameter estimation in dynamical systems via Statistical Learning: a
reinterpretation of Approximate Bayesian Computation applied to COVID-19
spread | We propose a robust parameter estimation method for dynamical systems based
on Statistical Learning techniques which aims to estimate a set of parameters
that well fit the dynamics in order to obtain robust evidences about the
qualitative behaviour of its trajectory. The method is quite general and
flexible, since it does not rely on any specific property of the dynamical
system, and represents a reinterpretation of Approximate Bayesian Computation
methods through the lens of Statistical Learning. The method is specially
useful for estimating parameters in epidemiological compartmental models in
order to obtain qualitative properties of a disease evolution. We apply it to
simulated and real data about COVID-19 spread in the US in order to evaluate
qualitatively its evolution over time, showing how one may assess the
effectiveness of measures implemented to slow the spread and some qualitative
features of the disease current and future evolution.
| stat |
Evaluating uncertainties in electrochemical impedance spectra of solid
oxide fuel cells | Electrochemical impedance spectroscopy (EIS) is a widely used tool for
characterization of fuel cells and other electrochemical conversion systems.
When applied to the on-line monitoring in the context of in-field applications,
the disturbances, drifts and sensor noise may cause severe distortions in the
evaluated spectra, especially in the low-frequency part. Failure to ignore the
random effects can result in misinterpreted spectra and, consequently, in
misleading diagnostic reasoning. This fact has not been often addressed in the
research so far. In this paper, we propose an approach to the quantification of
the spectral uncertainty, which relies on evaluating the uncertainty of the
equivalent circuit model (ECM). We apply the computationally efficient
variational Bayes (VB) method and compare the quality of the results with those
obtained with the Markov chain Monte Carlo (MCMC) algorithm. Namely, MCMC
algorithm returns accurate distributions of the estimated model parameters,
while VB approach provides the approximate distributions. By using simulated
and real data we show that approximate results provided by VB approach,
although slightly over-optimistic, are still close to the more realistic MCMC
estimates. A great advantage of the VB method for online monitoring is low
computational load, which is several orders of magnitude lower compared to
MCMC. The performance of VB algorithm is demonstrated on a case of ECM
parameters estimation in a 6 cell solid oxide fuel cell (SOFC) stack. The
complete numerical implementation for recreating the results can be found at
https://repo.ijs.si/lznidaric/variational-bayes-supplementary-material.
| stat |
Augmented Neural ODEs | We show that Neural Ordinary Differential Equations (ODEs) learn
representations that preserve the topology of the input space and prove that
this implies the existence of functions Neural ODEs cannot represent. To
address these limitations, we introduce Augmented Neural ODEs which, in
addition to being more expressive models, are empirically more stable,
generalize better and have a lower computational cost than Neural ODEs.
| stat |
Time Fused Coefficient SIR Model with Application to COVID-19 Epidemic
in the United States | In this paper, we propose a Susceptible-Infected-Removal (SIR) model with
time fused coefficients. In particular, our proposed model discovers the
underlying time homogeneity pattern for the SIR model's transmission rate and
removal rate via Bayesian shrinkage priors. MCMC sampling for the proposed
method is facilitated by the nimble package in R. Extensive simulation studies
are carried out to examine the empirical performance of the proposed methods.
We further apply the proposed methodology to analyze different levels of
COVID-19 data in the United States.
| stat |
Combinatorial Bayesian Optimization using the Graph Cartesian Product | This paper focuses on Bayesian Optimization (BO) for objectives on
combinatorial search spaces, including ordinal and categorical variables.
Despite the abundance of potential applications of Combinatorial BO, including
chipset configuration search and neural architecture search, only a handful of
methods have been proposed. We introduce COMBO, a new Gaussian Process (GP) BO.
COMBO quantifies "smoothness" of functions on combinatorial search spaces by
utilizing a combinatorial graph. The vertex set of the combinatorial graph
consists of all possible joint assignments of the variables, while edges are
constructed using the graph Cartesian product of the sub-graphs that represent
the individual variables. On this combinatorial graph, we propose an ARD
diffusion kernel with which the GP is able to model high-order interactions
between variables leading to better performance. Moreover, using the Horseshoe
prior for the scale parameter in the ARD diffusion kernel results in an
effective variable selection procedure, making COMBO suitable for high
dimensional problems. Computationally, in COMBO the graph Cartesian product
allows the Graph Fourier Transform calculation to scale linearly instead of
exponentially. We validate COMBO in a wide array of realistic benchmarks,
including weighted maximum satisfiability problems and neural architecture
search. COMBO outperforms consistently the latest state-of-the-art while
maintaining computational and statistical efficiency.
| stat |
Bayesian Paired-Comparison with the bpcs Package | This article introduces the bpcs R package (Bayesian Paired Comparison in
Stan) and the statistical models implemented in the package. The goal of this
package is to facilitate the use of Bayesian models for paired comparison data
in behavioral research. Historically, studies on preferences have relied on
Likert scale assessments and the frequentist approach to analyze the data. As
an alternative, this article proposes the use of Bayesian models for forced
choices assessments. The advantages of forced-choice assessments are that they
minimize common bias that Likert scales are susceptible to, they can increase
criterion validity, control for faking responses, and are quite useful to
assess preferences in studies with children and non-humans primates. Bayesian
data analyses are less sensitive to optional stopping, have better control of
type I error, have stronger evidence towards the null hypothesis, allows
propagation of uncertainties, includes prior information, and perform well when
handling models with many parameters and latent variables. The bpcs package
provides a consistent interface for R users and several functions to evaluate
the posterior distribution of all parameters, to estimate the posterior
distribution of any contest between items, and to obtain the posterior
distribution of the ranks. Three reanalyses of recent studies that used the
frequentist Bradley-Terry model are presented. These reanalyses are conducted
with the Bayesian models of the bpcs package and all the code used to fit the
models, generate the figures and the tables are available in the online
appendix.
| stat |
Statistical Network Analysis with Bergm | Recent advances in computational methods for intractable models have made
network data increasingly amenable to statistical analysis. Exponential random
graph models (ERGMs) emerged as one of the main families of models capable of
capturing the complex dependence structure of network data in a wide range of
applied contexts. The Bergm package for R has become a popular package to carry
out Bayesian parameter inference, missing data imputation, model selection and
goodness-of-fit diagnostics for ERGMs. Over the last few years, the package has
been considerably improved in terms of efficiency by adopting some of the
state-of-the-art Bayesian computational methods for doubly-intractable
distributions. Recently, version 5 of the package has been made available on
CRAN having undergone a substantial makeover, which has made it more accessible
and easy to use for practitioners. New functions include data augmentation
procedures based on the approximate exchange algorithm for dealing with missing
data, adjusted pseudo-likelihood and pseudo-posterior procedures, which allow
for fast approximate inference of the ERGM parameter posterior and model
evidence for networks on several thousands nodes.
| stat |
A local depth measure for general data | We introduce the Integrated Dual Local Depth which is a local depth measure
for data in a Banach space based on the use of one-dimensional projections. The
properties of a depth measure are analyzed under this setting and a proper
definition of local symmetry is given. Moreover, strong consistency results for
the local depth and also for the local depth regions are attained. Finally,
applications to descriptive data analysis and classification are analyzed,
making the special focus on multivariate functional data, where we obtain very
promising results.
| stat |
A practical, effective calculation of gamma difference distributions
with open data science tools | At present, there is still no officially accepted and extensively verified
implementation of computing the gamma difference distribution allowing unequal
shape parameters. We explore four computational ways of the gamma difference
distribution with the different shape parameters resulting from time series
kriging, a forecasting approach based on the best linear unbiased prediction,
and linear mixed models. The results of our numerical study, with emphasis on
using open data science tools, demonstrate that our open tool implemented in
high-performance Python(with Numba) is exponentially fast, highly accurate, and
very reliable. It combines numerical inversion of the characteristic function
and the trapezoidal rule with the double exponential oscillatory transformation
(DE quadrature). At the double 53-bit precision, our tool outperformed the
speed of the analytical computation based on Tricomi's $U(a, b, z)$ function in
CAS software (commercial Mathematica, open SageMath) by 1.5-2 orders. At the
precision of scientific numerical computational tools, it exceeded open SciPy,
NumPy, and commercial MATLAB 5-10 times. The potential future application of
our tool for a mixture of characteristic functions could open new possibilities
for fast data analysis based on exact probability distributions in areas like
multidimensional statistics, measurement uncertainty analysis in metrology as
well as in financial mathematics and risk analysis.
| stat |
Gaussian variational approximation for high-dimensional state space
models | Our article considers a Gaussian variational approximation of the posterior
density in a high-dimensional state space model. The variational parameters to
be optimized are the mean vector and the covariance matrix of the
approximation. The number of parameters in the covariance matrix grows as the
square of the number of model parameters, so it is necessary to find simple yet
effective parameterizations of the covariance structure when the number of
model parameters is large. We approximate the joint posterior distribution over
the high-dimensional state vectors by a dynamic factor model, having Markovian
time dependence and a factor covariance structure for the states. This gives a
reduced description of the dependence structure for the states, as well as a
temporal conditional independence structure similar to that in the true
posterior. The usefulness of the approach is illustrated for prediction in two
high-dimensional applications that are challenging for Markov chain Monte Carlo
sampling. The first is a spatio-temporal model for the spread of the Eurasian
Collared-Dove across North America; the second is a Wishart-based multivariate
stochastic volatility model for financial returns.
| stat |
Binary R Packages for Linux: Past, Present and Future | Pre-compiled binary packages provide a very convenient way of efficiently
distributing software that has been adopted by most Linux package management
systems. However, the heterogeneity of the Linux ecosystem, combined with the
growing number of R extensions available, poses a scalability problem. As a
result, efforts to bring binary R packages to Linux have been scattered, and
lack a proper mechanism to fully integrate them with R's package manager. This
work reviews past and present of binary distribution for Linux, and presents a
path forward by showcasing the `cran2copr' project, an RPM-based
proof-of-concept implementation of an automated scalable binary distribution
system with the capability of building, maintaining and distributing thousands
of packages, while providing a portable and extensible bridge to the system
package manager.
| stat |
Fast computation of latent correlations | Latent Gaussian copula models provide a powerful means to perform multi-view
data integration since these models can seamlessly express dependencies between
mixed variable types (binary, continuous, zero-inflated) via latent Gaussian
correlations. The estimation of these latent correlations, however, comes at
considerable computational cost, having prevented the routine use of these
models on high-dimensional data. Here, we propose a new computational approach
for estimating latent correlations via a hybrid multi-linear interpolation and
optimization scheme. Our approach speeds up the current state of the art
computation by several orders of magnitude, thus allowing fast computation of
latent Gaussian copula models even when the number of variables $p$ is large.
We provide theoretical guarantees for the approximation error of our numerical
scheme and support its excellent performance on simulated and real-world data.
We illustrate the practical advantages of our method on high-dimensional sparse
quantitative and relative abundance microbiome data as well as multi-view data
from The Cancer Genome Atlas Project. Our method is implemented in the R
package mixedCCA, available at https://github.com/irinagain/mixedCCA.
| stat |
Pricing service maintenance contracts using predictive analytics | As more manufacturers shift their focus from selling products to end
solutions, full-service maintenance contracts gain traction in the business
world. These contracts cover all maintenance related costs during a
predetermined horizon in exchange for a fixed service fee and relieve customers
from uncertain maintenance costs. To guarantee profitability, the service fees
should at least cover the expected costs during the contract horizon. As these
expected costs may depend on several machine-dependent characteristics, e.g.
operational environment, the service fees should also be differentiated based
on these characteristics. If not, customers that are less prone to high
maintenance costs will not buy into or renege on the contract. The latter can
lead to adverse selection and leave the service provider with a
maintenance-heavy portfolio, which may be detrimental to the profitability of
the service contracts. We contribute to the literature with a data-driven
tariff plan based on the calibration of predictive models that take into
account the different machine profiles. This conveys to the service provider
which machine profiles should be attracted at which price. We demonstrate the
advantage of a differentiated tariff plan and show how it better protects
against adverse selection.
| stat |
Symbolic Computation of Tight Causal Bounds | Causal inference involves making a set of assumptions about the nature of
things, defining a causal query, and attempting to find estimators of the query
based on the distribution of observed variables. When causal queries are not
identifiable from the observed data, it still may be possible to derive bounds
for these quantities in terms of the distribution of observed variables. We
develop and describe a general approach for computation of bounds, proving that
if the problem can be stated as a linear program, then the true global extrema
result in tight bounds. Building upon previous work in this area, we
characterize a class of problems that can always be stated as a linear
programming problem; we describe a general algorithm for constructing the
linear objective and constraints based on the causal model and the causal query
of interest. These problems therefore can be solved using a vertex enumeration
algorithm. We develop an R package implementing this algorithm with a user
friendly graphical interface using directed acyclic graphs, which only allows
for problems within this class to be depicted. We have implemented additional
features to help with interpreting and applying the bounds that we illustrate
in examples.
| stat |
Validation of Association | Recognizing, quantifying and visualizing associations between two variables
is increasingly important. This paper investigates how a new function-valued
measure of dependence, the quantile dependence function, can be used to
construct tests for independence and to provide an easily interpretable
diagnostic plot of existing departures from the null model. The dependence
function is designed to detect general dependence structure between variables
in quantiles of the joint distribution. It gives an insight into how the
dependence structures changes in different parts of the joint distribution. We
define new estimators of the dependence function, discuss some of their
properties, and apply them to construct new tests of independence. Numerical
evidence is given on the test's benefits against three recognized independence
tests introduced in the previous years. In real-data analysis, we illustrate
the use of our tests and the graphical presentation of the underlying
dependence structure.
| stat |
Sample Complexity Bounds for 1-bit Compressive Sensing and Binary Stable
Embeddings with Generative Priors | The goal of standard 1-bit compressive sensing is to accurately recover an
unknown sparse vector from binary-valued measurements, each indicating the sign
of a linear function of the vector. Motivated by recent advances in compressive
sensing with generative models, where a generative modeling assumption replaces
the usual sparsity assumption, we study the problem of 1-bit compressive
sensing with generative models. We first consider noiseless 1-bit measurements,
and provide sample complexity bounds for approximate recovery under
i.i.d.~Gaussian measurements and a Lipschitz continuous generative prior, as
well as a near-matching algorithm-independent lower bound. Moreover, we
demonstrate that the Binary $\epsilon$-Stable Embedding property, which
characterizes the robustness of the reconstruction to measurement errors and
noise, also holds for 1-bit compressive sensing with Lipschitz continuous
generative models with sufficiently many Gaussian measurements. In addition, we
apply our results to neural network generative models, and provide a
proof-of-concept numerical experiment demonstrating significant improvements
over sparsity-based approaches.
| stat |
The Zoltar forecast archive: a tool to facilitate standardization and
storage of interdisciplinary prediction research | Forecasting has emerged as an important component of informed, data-driven
decision-making in a wide array of fields. We introduce a new data model for
probabilistic predictions that encompasses a wide range of forecasting
settings. This framework clearly defines the constituent parts of a
probabilistic forecast and proposes one approach for representing these data
elements. The data model is implemented in Zoltar, a new software application
that stores forecasts using the data model and provides standardized API access
to the data. In one real-time case study, an instance of the Zoltar web
application was used to store, provide access to, and evaluate real-time
forecast data on the order of 10$^7$ rows, provided by over 20 international
research teams from academia and industry making forecasts of the COVID-19
outbreak in the US. Tools and data infrastructure for probabilistic forecasts,
such as those introduced here, will play an increasingly important role in
ensuring that future forecasting research adheres to a strict set of rigorous
and reproducible standards.
| stat |
Hospital Capacity Planning Using Discrete Event Simulation Under Special
Consideration of the COVID-19 Pandemic | We present a resource-planning tool for hospitals under special consideration
of the COVID-19 pandemic, called babsim.hospital. It provides many advantages
for crisis teams, e.g., comparison with their own local planning, simulation of
local events, simulation of several scenarios (worst / best case). There are
benefits for medical professionals, e.g, analysis of the pandemic at local,
regional, state and federal level, the consideration of special risk groups,
tools for validating the length of stays and transition probabilities. Finally,
there are potential advantages for administration, management, e.g., assessment
of the situation of individual hospitals taking local events into account,
consideration of relevant resources such as beds, ventilators, rooms,
protective clothing, and personnel planning, e.g., medical and nursing staff.
babsim.hospital combines simulation, optimization, statistics, and artificial
intelligence processes in a very efficient way. The core is a discrete,
event-based simulation model.
| stat |
Correcting misclassification errors in crowdsourced ecological data: A
Bayesian perspective | Many research domains use data elicited from "citizen scientists" when a
direct measure of a process is expensive or infeasible. However, participants
may report incorrect estimates or classifications due to their lack of skill.
We demonstrate how Bayesian hierarchical models can be used to learn about
latent variables of interest, while accounting for the participants' abilities.
The model is described in the context of an ecological application that
involves crowdsourced classifications of georeferenced coral-reef images from
the Great Barrier Reef, Australia. The latent variable of interest is the
proportion of coral cover, which is a common indicator of coral reef health.
The participants' abilities are expressed in terms of sensitivity and
specificity of a correctly classified set of points on the images. The model
also incorporates a spatial component, which allows prediction of the latent
variable in locations that have not been surveyed. We show that the model
outperforms traditional weighted-regression approaches used to account for
uncertainty in citizen science data. Our approach produces more accurate
regression coefficients and provides a better characterization of the latent
process of interest. This new method is implemented in the probabilistic
programming language Stan and can be applied to a wide number of problems that
rely on uncertain citizen science data.
| stat |
On estimating the size of overcoverage with the latent class model. A
critique of the paper "Population Size Estimation Using Multiple Incomplete
Lists with Overcoverage" by di Cecco, di Zio, Filipponi and Rocchetti (2018,
JOS 34 557-572) | We read with interest the article by di Cecco et al. (2018), but have
reservations about the usefulness of the latent class model specifically for
estimating overcoverage. In particular, we question the interpretation of the
parameters of the fitted latent class model.
| stat |
Transmission Matrix Inference via Pseudolikelihood Decimation | One of the biggest challenges in the field of biomedical imaging is the
comprehension and the exploitation of the photon scattering through disordered
media. Many studies have pursued the solution to this puzzle, achieving
light-focusing control or reconstructing images in complex media. In the
present work, we investigate how statistical inference helps the calculation of
the transmission matrix in a complex scrambling environment, enabling its usage
like a normal optical element. We convert a linear input-output transmission
problem into a statistical formulation based on pseudolikelihood maximization,
learning the coupling matrix via random sampling of intensity realizations. Our
aim is to uncover insights from the scattering problem, encouraging the
development of novel imaging techniques for better medical investigations,
borrowing a number of statistical tools from spin-glass theory.
| stat |
MLSolv-A: A Novel Machine Learning-Based Prediction of Solvation Free
Energies from Pairwise Atomistic Interactions | Recent advances in machine learning and their applications have lead to the
development of diverse structure-property relationship models for crucial
chemical properties, and the solvation free energy is one of them. Here, we
introduce a novel ML-based solvation model, which calculates the solvation
energy from pairwise atomistic interactions. The novelty of the proposed model
consists of a simple architecture: two encoding functions extract atomic
feature vectors from the given chemical structure, while the inner product
between two atomistic features calculates their interactions. The results on
6,493 experimental measurements achieve outstanding performance and
transferability for enlarging training data due to its solvent-non-specific
nature. Analysis of the interaction map shows there is a great potential that
our model reproduces group contributions on the solvation energy, which makes
us believe that the model not only provides the predicted target property but
also gives us more detailed physicochemical insights.
| stat |
Upper Trust Bound Feasibility Criterion for Mixed Constrained Bayesian
Optimization with Application to Aircraft Design | Bayesian optimization methods have been successfully applied to black box
optimization problems that are expensive to evaluate. In this paper, we adapt
the so-called super effcient global optimization algorithm to solve more
accurately mixed constrained problems. The proposed approach handles
constraints by means of upper trust bound, the latter encourages exploration of
the feasible domain by combining the mean prediction and the associated
uncertainty function given by the Gaussian processes. On top of that, a
refinement procedure, based on a learning rate criterion, is introduced to
enhance the exploitation and exploration trade-off. We show the good potential
of the approach on a set of numerical experiments. Finally, we present an
application to conceptual aircraft configuration upon which we show the
superiority of the proposed approach compared to a set of the state-of-the-art
black box optimization solvers. Keywords: Global Optimization, Mixed
Constrained Optimization, Black box optimization, Bayesian Optimization,
Gaussian Process.
| stat |
The role of intrinsic dimension in high-resolution player tracking data
-- Insights in basketball | A new range of statistical analysis has emerged in sports after the
introduction of the high-resolution player tracking technology, specifically in
basketball. However, this high dimensional data is often challenging for
statistical inference and decision making. In this article, we employ Hidalgo,
a state-of-the-art Bayesian mixture model that allows the estimation of
heterogeneous intrinsic dimensions (ID) within a dataset and propose some
theoretical enhancements. ID results can be interpreted as indicators of
variability and complexity of basketball plays and games. This technique allows
classification and clustering of NBA basketball player's movement and shot
charts data. Analyzing movement data, Hidalgo identifies key stages of
offensive actions such as creating space for passing, preparation/shooting and
following through. We found that the ID value spikes reaching a peak between 4
and 8 seconds in the offensive part of the court after which it declines. In
shot charts, we obtained groups of shots that produce substantially higher and
lower successes. Overall, game-winners tend to have a larger intrinsic
dimension which is an indication of more unpredictability and unique shot
placements. Similarly, we found higher ID values in plays when the score margin
is small compared to large margin ones. These outcomes could be exploited by
coaches to obtain better offensive/defensive results.
| stat |
CRAD: Clustering with Robust Autocuts and Depth | We develop a new density-based clustering algorithm named CRAD which is based
on a new neighbor searching function with a robust data depth as the
dissimilarity measure. Our experiments prove that the new CRAD is highly
competitive at detecting clusters with varying densities, compared with the
existing algorithms such as DBSCAN, OPTICS and DBCA. Furthermore, a new
effective parameter selection procedure is developed to select the optimal
underlying parameter in the real-world clustering, when the ground truth is
unknown. Lastly, we suggest a new clustering framework that extends CRAD from
spatial data clustering to time series clustering without a-priori knowledge of
the true number of clusters. The performance of CRAD is evaluated through
extensive experimental studies.
| stat |
Causal Discovery with Unobserved Confounding and non-Gaussian Data | We consider the problem of recovering causal structure from multivariate
observational data. We assume that the data arise from a linear structural
equation model (SEM) in which the idiosyncratic errors are allowed to be
dependent in order to capture possible latent confounding. Each SEM can be
represented by a graph where vertices represent observed variables, directed
edges represent direct causal effects, and bidirected edges represent
dependence among error terms. Specifically, we assume that the true model
corresponds to a bow-free acyclic path diagram, i.e., a graph that has at most
one edge between any pair of nodes and is acyclic in the directed part. We show
that when the errors are non-Gaussian, the exact causal structure encoded by
such a graph, and not merely an equivalence class, can be consistently
recovered from observational data. The Bow-free Acylic Non-Gaussian (BANG)
method we propose for this purpose uses estimates of suitable moments, but, in
contrast to previous results, does not require specifying the number of latent
variables a priori. We illustrate the effectiveness of BANG in simulations and
an application to an ecology data set.
| stat |
Online detection of cascading change-points | We propose an online detection procedure for cascading failures in the
network from sequential data, which can be modeled as multiple correlated
change-points happening during a short period. We consider a temporal diffusion
network model to capture the temporal dynamic structure of multiple
change-points and develop a sequential Shewhart procedure based on the
generalized likelihood ratio statistics based on the diffusion network model
assuming unknown post-change distribution parameters. We also tackle the
computational complexity posed by the unknown propagation. Numerical
experiments demonstrate the good performance for detecting cascade failures.
| stat |
How have German University Tuition Fees Affected Enrollment Rates:
Robust Model Selection and Design-based Inference in High-Dimensions | We use official data for all 16 federal German states to study the causal
effect of a flat 1000 Euro state-dependent university tuition fee on the
enrollment behavior of students during the years 2006-2014. In particular, we
show how the variation in the introduction scheme across states and times can
be exploited to identify the federal average causal effect of tuition fees by
controlling for a large amount of potentially influencing attributes for state
heterogeneity. We suggest a stability post-double selection methodology to
robustly determine the causal effect across types in the transparently modeled
unknown response components. The proposed stability resampling scheme in the
two LASSO selection steps efficiently mitigates the risk of model
underspecification and thus biased effects when the tuition fee policy decision
also depends on relevant variables for the state enrollment rates. Correct
inference for the full cross-section state population in the sample requires
adequate design -- rather than sampling-based standard errors. With the
data-driven model selection and explicit control for spatial cross-effects we
detect that tuition fees induce substantial migration effects where the
mobility occurs both from fee but also from non-fee states suggesting also a
general movement for quality. Overall, we find a significant negative impact of
up to 4.5 percentage points of fees on student enrollment. This is in contrast
to plain one-step LASSO or previous empirical studies with full fixed effects
linear panel regressions which generally underestimate the size and get an only
insignificant effect.
| stat |
Ambiguity in Sequential Data: Predicting Uncertain Futures with
Recurrent Models | Ambiguity is inherently present in many machine learning tasks, but
especially for sequential models seldom accounted for, as most only output a
single prediction. In this work we propose an extension of the Multiple
Hypothesis Prediction (MHP) model to handle ambiguous predictions with
sequential data, which is of special importance, as often multiple futures are
equally likely. Our approach can be applied to the most common recurrent
architectures and can be used with any loss function. Additionally, we
introduce a novel metric for ambiguous problems, which is better suited to
account for uncertainties and coincides with our intuitive understanding of
correctness in the presence of multiple labels. We test our method on several
experiments and across diverse tasks dealing with time series data, such as
trajectory forecasting and maneuver prediction, achieving promising results.
| stat |
Personalized Treatment for Coronary Artery Disease Patients: A Machine
Learning Approach | Current clinical practice guidelines for managing Coronary Artery Disease
(CAD) account for general cardiovascular risk factors. However, they do not
present a framework that considers personalized patient-specific
characteristics. Using the electronic health records of 21,460 patients, we
created data-driven models for personalized CAD management that significantly
improve health outcomes relative to the standard of care. We develop binary
classifiers to detect whether a patient will experience an adverse event due to
CAD within a 10-year time frame. Combining the patients' medical history and
clinical examination results, we achieve 81.5% AUC. For each treatment, we also
create a series of regression models that are based on different supervised
machine learning algorithms. We are able to estimate with average R squared =
0.801 the time from diagnosis to a potential adverse event (TAE) and gain
accurate approximations of the counterfactual treatment effects. Leveraging
combinations of these models, we present ML4CAD, a novel personalized
prescriptive algorithm. Considering the recommendations of multiple predictive
models at once, ML4CAD identifies for every patient the therapy with the best
expected outcome using a voting mechanism. We evaluate its performance by
measuring the prescription effectiveness and robustness under alternative
ground truths. We show that our methodology improves the expected TAE upon the
current baseline by 24.11%, increasing it from 4.56 to 5.66 years. The
algorithm performs particularly well for the male (24.3% improvement) and
Hispanic (58.41% improvement) subpopulations. Finally, we create an interactive
interface, providing physicians with an intuitive, accurate, readily
implementable, and effective tool.
| stat |
Augmented pseudo-marginal Metropolis-Hastings for partially observed
diffusion processes | We consider the problem of inference for nonlinear, multivariate diffusion
processes, satisfying It\^o stochastic differential equations (SDEs), using
data at discrete times that may be incomplete and subject to measurement error.
Our starting point is a state-of-the-art correlated pseudo-marginal
Metropolis-Hastings scheme, that uses correlated particle filters to induce
strong and positive correlation between successive marginal likelihood
estimates. However, unless the measurement error or the dimension of the SDE is
small, correlation can be eroded by the resampling steps in the particle
filter. We therefore propose a novel augmentation scheme, that allows for
conditioning on values of the latent process at the observation times,
completely avoiding the need for resampling steps. We integrate over the
uncertainty at the observation times with an additional Gibbs step. Connections
between the resulting pseudo-marginal scheme and existing inference schemes for
diffusion processes are made. The methodology is applied in three examples of
increasing complexity. We find that our approach offers substantial increases
in overall efficiency, compared to competing methods.
| stat |
A Nonparametric Approach to High-dimensional k-sample Comparison
Problems | High-dimensional k-sample comparison is a common applied problem. We
construct a class of easy-to-implement nonparametric distribution-free tests
based on new tools and unexplored connections with spectral graph theory. The
test is shown to possess various desirable properties along with a
characteristic exploratory flavor that has practical consequences. The
numerical examples show that our method works surprisingly well under a broad
range of realistic situations.
| stat |
Estimating heterogeneous treatment effects in nonstationary time series
with state-space models | Randomized trials and observational studies, more often than not, run over a
certain period of time. The treatment effect evolves during this period which
provides crucial insights into the treatment response and the long-term
effects. Many conventional methods for estimating treatment effects are limited
to the i.i.d. setting and are not suited for inferring the time dynamics of the
treatment effect. The time series encountered in these settings are highly
informative but often nonstationary due to the changing effects of treatment.
This increases the difficulty, since stationarity, a common assumption in time
series analysis, cannot be reasonably assumed. Another challenge is the
heterogeneity of the treatment effect when the treatment affects units
differently. The task of estimating heterogeneous treatment effects from
nonstationary and, in particular, interventional time series is highly relevant
but has remained unexplored yet. We propose Causal Transfer, a method which
combines regression to adjust for confounding with time series modelling to
learn the effect of the treatment and how it evolves over time. Causal Transfer
does not assume the data to be stationary and can be applied to randomized
trials and observational studies in which treatment is confounded. Causal
Transfer adjusts the effect for possible confounders and transfers the learned
effect to other time series and, thereby, estimates various forms of treatment
effects, such as the average treatment effect (ATE) or the conditional average
treatment effect (CATE). By learning the time dynamics of the effect, Causal
Transfer can also predict the treatment effect for unobserved future time
points and determine the long-term consequences of treatment.
| stat |
Multi-level Monte Carlo Variational Inference | We propose a variance reduction framework for variational inference using the
multi-level Monte Carlo (MLMC) method. The proposed framework "recycles"
parameters obtained from past update history in optimization and can be
compatible with reparameterized gradient estimators. Our framework provides a
novel optimization algorithm based on the stochastic gradient method and
adaptively estimates the sample size for stochastic gradient estimation per
level according to the ratio of the variance and computation cost in each
iteration. We also analyze the convergence of the gradient norm in the
stochastic gradient method, the scale of the gradient estimator's variance, and
the estimator's quality in each optimization step on the basis of the
\textit{signal-to-noise} ratio. Finally, we experimentally evaluate the
proposed method by comparing it with baseline methods on several benchmark data
sets. The results confirm that the proposed method achieves faster convergence
and reduces the variance of the gradient estimator compared with the other
methods.
| stat |
Robustness to Adversarial Perturbations in Learning from Incomplete Data | What is the role of unlabeled data in an inference problem, when the presumed
underlying distribution is adversarially perturbed? To provide a concrete
answer to this question, this paper unifies two major learning frameworks:
Semi-Supervised Learning (SSL) and Distributionally Robust Learning (DRL). We
develop a generalization theory for our framework based on a number of novel
complexity measures, such as an adversarial extension of Rademacher complexity
and its semi-supervised analogue. Moreover, our analysis is able to quantify
the role of unlabeled data in the generalization under a more general condition
compared to the existing theoretical works in SSL. Based on our framework, we
also present a hybrid of DRL and EM algorithms that has a guaranteed
convergence rate. When implemented with deep neural networks, our method shows
a comparable performance to those of the state-of-the-art on a number of
real-world benchmark datasets.
| stat |
Permutation inference in factorial survival designs with the CASANOVA | We propose inference procedures for general nonparametric factorial survival
designs with possibly right-censored data. Similar to additive Aalen models,
null hypotheses are formulated in terms of cumulative hazards. Thereby,
deviations are measured in terms of quadratic forms in Nelson-Aalen-type
integrals. Different to existing approaches this allows to work without
restrictive model assumptions as proportional hazards. In particular, crossing
survival or hazard curves can be detected without a significant loss of power.
For a distribution-free application of the method, a permutation strategy is
suggested. The resulting procedures' asymptotic validity as well as their
consistency are proven and their small sample performances are analyzed in
extensive simulations. Their applicability is finally illustrated by analyzing
an oncology data set.
| stat |
A slice tour for finding hollowness in high-dimensional data | Taking projections of high-dimensional data is a common analytical and
visualisation technique in statistics for working with high-dimensional
problems. Sectioning, or slicing, through high dimensions is less common, but
can be useful for visualising data with concavities, or non-linear structure.
It is associated with conditional distributions in statistics, and also linked
brushing between plots in interactive data visualisation. This short technical
note describes a simple approach for slicing in the orthogonal space of
projections obtained when running a tour, thus presenting the viewer with an
interpolated sequence of sliced projections. The method has been implemented in
R as an extension to the tourr package, and can be used to explore for concave
and non-linear structures in multivariate distributions.
| stat |
Comparing distributions: $\ell_1$ geometry improves kernel two-sample
testing | Are two sets of observations drawn from the same distribution? This problem
is a two-sample test. Kernel methods lead to many appealing properties. Indeed
state-of-the-art approaches use the $L^2$ distance between kernel-based
distribution representatives to derive their test statistics. Here, we show
that $L^p$ distances (with $p\geq 1$) between these distribution
representatives give metrics on the space of distributions that are
well-behaved to detect differences between distributions as they metrize the
weak convergence. Moreover, for analytic kernels, we show that the $L^1$
geometry gives improved testing power for scalable computational procedures.
Specifically, we derive a finite dimensional approximation of the metric given
as the $\ell_1$ norm of a vector which captures differences of expectations of
analytic functions evaluated at spatial locations or frequencies (i.e,
features). The features can be chosen to maximize the differences of the
distributions and give interpretable indications of how they differs. Using an
$\ell_1$ norm gives better detection because differences between
representatives are dense as we use analytic kernels (non-zero almost
everywhere). The tests are consistent, while much faster than state-of-the-art
quadratic-time kernel-based tests. Experiments on artificial and real-world
problems demonstrate improved power/time tradeoff than the state of the art,
based on $\ell_2$ norms, and in some cases, better outright power than even the
most expensive quadratic-time tests.
| stat |
Assurance for sample size determination in reliability demonstration
testing | Manufacturers are required to demonstrate products meet reliability targets.
A typical way to achieve this is with reliability demonstration tests (RDTs),
in which a number of products are put on test and the test is passed if a
target reliability is achieved. There are various methods for determining the
sample size for RDTs, typically based on the power of a hypothesis test
following the RDT or risk criteria. Bayesian risk criteria approaches can
conflate the choice of sample size and the analysis to be undertaken once the
test has been conducted and rely on the specification of somewhat artificial
acceptable and rejectable reliability levels. In this paper we offer an
alternative approach to sample size determination based on the idea of
assurance. This approach chooses the sample size to answer provide a certain
probability that the RDT will result in a successful outcome. It separates the
design and analysis of the RDT, allowing different priors for each. We develop
the assurance approach for sample size calculations in RDTs for binomial and
Weibull likelihoods and propose appropriate prior distributions for the design
and analysis of the test. In each case, we illustrate the approach with an
example based on real data.
| stat |
Optimization of Distribution Network Configuration for Pediatric
Vaccines using Chance Constraint Programming | Millions of young people are not immunized in low- and middle-income (LMI)
countries because of low vaccine availability resulting from inefficiencies in
cold supply chains. We create supply chain network design and distribution
models to address the unique characteristics and challenges facing vaccine
supply chains in LMI countries. The models capture the uncertainties of demand
for vaccinations and the resulting impacts on immunization, the unique
challenges of vaccine administration (such as open vial wastage), the
interactions between technological improvements of vaccines and immunizations,
and the trade-offs between immunization coverage rates and available resources.
The objective is to maximize both the percentage of fully immunized children
and the vaccine availability in clinics. Our research examines how these two
metrics are affected by three factors: number of tiers in the supply chain,
vaccine vial size, and new vaccine technologies. We tested the model using
Niger's Expanded Program on Immunization, which is sponsored by the World
Health Organization. We make many observations and recommendations to help LMI
countries increase their immunization coverage.
| stat |
Fast Spatial Autocorrelation | Physical or geographic location proves to be an important feature in many
data science models, because many diverse natural and social phenomenon have a
spatial component. Spatial autocorrelation measures the extent to which locally
adjacent observations of the same phenomenon are correlated. Although
statistics like Moran's $I$ and Geary's $C$ are widely used to measure spatial
autocorrelation, they are slow: all popular methods run in $\Omega(n^2)$ time,
rendering them unusable for large data sets, or long time-courses with moderate
numbers of points. We propose a new $S_A$ statistic based on the notion that
the variance observed when merging pairs of nearby clusters should increase
slowly for spatially autocorrelated variables. We give a linear-time algorithm
to calculate $S_A$ for a variable with an input agglomeration order (available
at https://github.com/aamgalan/spatial_autocorrelation). For a typical dataset
of $n \approx 63,000$ points, our $S_A$ autocorrelation measure can be computed
in 1 second, versus 2 hours or more for Moran's $I$ and Geary's $C$. Through
simulation studies, we demonstrate that $S_A$ identifies spatial correlations
in variables generated with spatially-dependent model half an order of
magnitude earlier than either Moran's $I$ or Geary's $C$. Finally, we prove
several theoretical properties of $S_A$: namely that it behaves as a true
correlation statistic, and is invariant under addition or multiplication by a
constant.
| stat |
Machine learning for subgroup discovery under treatment effect | In many practical tasks it is needed to estimate an effect of treatment on
individual level. For example, in medicine it is essential to determine the
patients that would benefit from a certain medicament. In marketing, knowing
the persons that are likely to buy a new product would reduce the amount of
spam. In this chapter, we review the methods to estimate an individual
treatment effect from a randomized trial, i.e., an experiment when a part of
individuals receives a new treatment, while the others do not. Finally, it is
shown that new efficient methods are needed in this domain.
| stat |
Modularity maximisation for graphons | Networks are a widely-used tool to investigate the large-scale connectivity
structure in complex systems and graphons have been proposed as an infinite
size limit of dense networks. The detection of communities or other meso-scale
structures is a prominent topic in network science as it allows the
identification of functional building blocks in complex systems. When such
building blocks may be present in graphons is an open question. In this paper,
we define a graphon-modularity and demonstrate that it can be maximised to
detect communities in graphons. We then investigate specific synthetic graphons
and show that they may show a wide range of different community structures. We
also reformulate the graphon-modularity maximisation as a continuous
optimisation problem and so prove the optimal community structure or lack
thereof for some graphons, something that is usually not possible for networks.
Furthermore, we demonstrate that estimating a graphon from network data as an
intermediate step can improve the detection of communities, in comparison with
exclusively maximising the modularity of the network. While the choice of
graphon-estimator may strongly influence the accord between the community
structure of a network and its estimated graphon, we find that there is a
substantial overlap if an appropriate estimator is used. Our study demonstrates
that community detection for graphons is possible and may serve as a
privacy-preserving way to cluster network data.
| stat |
The hidden waves in the ECG uncovered: a sound automated interpretation
method | A novel approach for analysing cardiac rhythm data is presented in this
paper. Heartbeats are decomposed into the five fundamental $P$, $Q$, $R$, $S$
and $T$ waves plus an error term to account for artefacts in the data which
provides a meaningful, physical interpretation of the heart's electric system.
The morphology of each wave is concisely described using four parameters that
allow to all the different patterns in heartbeats be characterized and thus
differentiated
This multi-purpose approach solves such questions as the extraction of
interpretable features, the detection of the fiducial marks of the fundamental
waves, or the generation of synthetic data and the denoising of signals. Yet,
the greatest benefit from this new discovery will be the automatic diagnosis of
heart anomalies as well as other clinical uses with great advantages compared
to the rigid, vulnerable and black box machine learning procedures, widely used
in medical devices.
The paper shows the enormous potential of the method in practice;
specifically, the capability to discriminate subjects, characterize
morphologies and detect the fiducial marks (reference points) are validated
numerically using simulated and real data, thus proving that it outperforms its
competitors.
| stat |
Beyond Binomial and Negative Binomial: Adaptation in Bernoulli Parameter
Estimation | Estimating the parameter of a Bernoulli process arises in many applications,
including photon-efficient active imaging where each illumination period is
regarded as a single Bernoulli trial. Motivated by acquisition efficiency when
multiple Bernoulli processes are of interest, we formulate the allocation of
trials under a constraint on the mean as an optimal resource allocation
problem. An oracle-aided trial allocation demonstrates that there can be a
significant advantage from varying the allocation for different processes and
inspires a simple trial allocation gain quantity. Motivated by realizing this
gain without an oracle, we present a trellis-based framework for representing
and optimizing stopping rules. Considering the convenient case of Beta priors,
three implementable stopping rules with similar performances are explored, and
the simplest of these is shown to asymptotically achieve the oracle-aided trial
allocation. These approaches are further extended to estimating functions of a
Bernoulli parameter. In simulations inspired by realistic active imaging
scenarios, we demonstrate significant mean-squared error improvements: up to
4.36 dB for the estimation of p and up to 1.80 dB for the estimation of log p.
| stat |
Discovering Multiple Phases of Dynamics by Dissecting Multivariate Time
Series | We proposed a data-driven approach to dissect multivariate time series in
order to discover multiple phases underlying dynamics of complex systems. This
computing approach is developed as a multiple-dimension version of Hierarchical
Factor Segmentation(HFS) technique. This expanded approach proposes a
systematic protocol of choosing various extreme events in multi-dimensional
space. Upon each chosen event, an empirical distribution of event-recurrence,
or waiting time between the excursions, is fitted by a geometric distribution
with time-varying parameters. Iterative fittings are performed across all
chosen events. We then collect and summarize the local recurrent patterns into
a global dynamic mechanism. Clustering is applied for partitioning the whole
time period into alternating segments, in which variables are identically
distributed. Feature weighting techniques are also considered to compensate for
some drawbacks of clustering. Our simulation results show that this expanded
approach can even detect systematic differences when the joint distribution
varies. In real data experiments, we analyze the relationship from returns,
trading volume, and transaction number of a single, as well as of multiple
stocks in S&P500. We can successfully not only map out volatile periods but
also provide potential associative links between stocks.
| stat |
Explicit Regularisation in Gaussian Noise Injections | We study the regularisation induced in neural networks by Gaussian noise
injections (GNIs). Though such injections have been extensively studied when
applied to data, there have been few studies on understanding the regularising
effect they induce when applied to network activations. Here we derive the
explicit regulariser of GNIs, obtained by marginalising out the injected noise,
and show that it penalises functions with high-frequency components in the
Fourier domain; particularly in layers closer to a neural network's output. We
show analytically and empirically that such regularisation produces calibrated
classifiers with large classification margins.
| stat |
Selecting time-series hyperparameters with the artificial jackknife | This article proposes a generalisation of the delete-$d$ jackknife to solve
hyperparameter selection problems for time series. This novel technique is
compatible with dependent data since it substitutes the jackknife removal step
with a fictitious deletion, wherein observed datapoints are replaced with
artificial missing values. In order to emphasise this point, I called this
methodology artificial delete-$d$ jackknife. As an illustration, it is used to
regulate vector autoregressions with an elastic-net penalty on the
coefficients. A software implementation, ElasticNetVAR.jl, is available on
GitHub.
| stat |
Bayesian cumulative shrinkage for infinite factorizations | There is a wide variety of models in which the dimension of the parameter
space is unknown. For example, in factor analysis the number of latent factors
is typically not known and has to be inferred from the observed data. Although
classical shrinkage priors are useful in these contexts, increasing shrinkage
priors can provide a more effective option, which progressively penalizes
expansions with growing complexity. In this article we propose a novel
increasing shrinkage prior, named the cumulative shrinkage process, for the
parameters controlling the dimension in over-complete formulations. Our
construction has broad applicability, simple interpretation, and is based on a
sequence of spike and slab distributions which assign increasing mass to the
spike as model complexity grows. Using factor analysis as an illustrative
example, we show that this formulation has theoretical and practical advantages
over current competitors, including an improved ability to recover the model
dimension. An adaptive Markov chain Monte Carlo algorithm is proposed, and the
methods are evaluated in simulation studies and applied to personality traits
data.
| stat |
General Fair Empirical Risk Minimization | We tackle the problem of algorithmic fairness, where the goal is to avoid the
unfairly influence of sensitive information, in the general context of
regression with possible continuous sensitive attributes. We extend the
framework of fair empirical risk minimization to this general scenario,
covering in this way the whole standard supervised learning setting. Our
generalized fairness measure reduces to well known notions of fairness
available in literature. We derive learning guarantees for our method, that
imply in particular its statistical consistency, both in terms of the risk and
the fairness measure. We then specialize our approach to kernel methods and
propose a convex fair estimator in that setting. We test the estimator on a
commonly used benchmark dataset (Communities and Crime) and on a new dataset
collected at the University of Genova, containing the information of the
academic career of five thousand students. The latter dataset provides a
challenging real case scenario of unfair behaviour of standard regression
methods that benefits from our methodology. The experimental results show that
our estimator is effective at mitigating the trade-off between accuracy and
fairness requirements.
| stat |
An Easy-to-Implement Hierarchical Standardization for Variable Selection
Under Strong Heredity Constraint | For many practical problems, the regression models follow the strong heredity
property (also known as the marginality), which means they include parent main
effects when a second-order effect is present. Existing methods rely mostly on
special penalty functions or algorithms to enforce the strong heredity in
variable selection. We propose a novel hierarchical standardization procedure
to maintain strong heredity in variable selection. Our method is effortless to
implement and is applicable to any variable selection method for any type of
regression. The performance of the hierarchical standardization is comparable
to that of the regular standardization. We also provide robustness checks and
real data analysis to illustrate the merits of our method.
| stat |
Matrix Linear Discriminant Analysis | We propose a novel linear discriminant analysis approach for the
classification of high-dimensional matrix-valued data that commonly arises from
imaging studies. Motivated by the equivalence of the conventional linear
discriminant analysis and the ordinary least squares, we consider an efficient
nuclear norm penalized regression that encourages a low-rank structure.
Theoretical properties including a non-asymptotic risk bound and a rank
consistency result are established. Simulation studies and an application to
electroencephalography data show the superior performance of the proposed
method over the existing approaches.
| stat |
Integrative Learning for Population of Dynamic Networks with Covariates | Although there is a rapidly growing literature on dynamic connectivity
methods, the primary focus has been on separate network estimation for each
individual, which fails to leverage common patterns of information. We propose
novel graph-theoretic approaches for estimating a population of dynamic
networks that are able to borrow information across multiple heterogeneous
samples in an unsupervised manner and guided by covariate information.
Specifically, we develop a Bayesian product mixture model that imposes
independent mixture priors at each time scan and uses covariates to model the
mixture weights, which results in time-varying clusters of samples designed to
pool information. The computation is carried out using an efficient
Expectation-Maximization algorithm. Extensive simulation studies illustrate
sharp gains in recovering the true dynamic network over existing dynamic
connectivity methods. An analysis of fMRI block task data with behavioral
interventions reveal sub-groups of individuals having similar dynamic
connectivity, and identifies intervention-related dynamic network changes that
are concentrated in biologically interpretable brain regions. In contrast,
existing dynamic connectivity approaches are able to detect minimal or no
changes in connectivity over time, which seems biologically unrealistic and
highlights the challenges resulting from the inability to systematically borrow
information across samples.
| stat |
Revisiting the Gelman-Rubin Diagnostic | Gelman and Rubin's (1992) convergence diagnostic is one of the most popular
methods for terminating a Markov chain Monte Carlo (MCMC) sampler. Since the
seminal paper, researchers have developed sophisticated methods for estimating
variance of Monte Carlo averages. We show that these estimators find immediate
use in the Gelman-Rubin statistic, a connection not previously established in
the literature. We incorporate these estimators to upgrade both the univariate
and multivariate Gelman-Rubin statistics, leading to improved stability in MCMC
termination time. An immediate advantage is that our new Gelman-Rubin statistic
can be calculated for a single chain. In addition, we establish a one-to-one
relationship between the Gelman-Rubin statistic and effective sample size.
Leveraging this relationship, we develop a principled termination criterion for
the Gelman-Rubin statistic. Finally, we demonstrate the utility of our improved
diagnostic via examples.
| stat |
Present criteria for prophylactic ICD implantation: Insights from the
EU-CERT-ICD (Comparative Effectiveness Research to Assess the Use of Primary
ProphylacTic Implantable Cardioverter Defibrillators in EUrope) project | BACKGROUND. The clinical effectiveness of primary prevention implantable
cardioverter defibrillator (ICD) therapy is under debate. It is urgently needed
to better identify patients who benefit from prophylactic ICD therapy. The
EUropean Comparative Effectiveness Research to Assess the Use of Primary
ProphylacTic Implantable Cardioverter Defibrillators (EU-CERT-ICD) completed in
2019 will assess this issue.
SUMMARY. The EU-CERT-ICD is a prospective investigator-initiated
non-randomized, controlled, multicenter observational cohort study done in 44
centers across 15 European countries. A total of 2327 patients with heart
failure due to ischemic heart disease or dilated cardiomyopathy indicated for
primary prophylactic ICD implantation were recruited between 2014 and 2018
(>1500 patients at first ICD implantation, >750 patients non-randomized non-ICD
control group). The primary endpoint was all-cause mortality, first appropriate
shock was co-primary endpoint. At baseline, all patients underwent 12-lead ECG
and Holter-ECG analysis using multiple advanced methods for risk stratification
as well as documentation of clinical characteristics and laboratory values. The
EU-CERT-ICD data will provide much needed information on the survival benefit
of preventive ICD therapy and expand on previous prospective risk
stratification studies which showed very good applicability of clinical
parameters and advanced risk stratifiers in order to define patient subgroups
with above or below average ICD benefit.
CONCLUSION. The EU-CERT-ICD study will provide new and current data about
effectiveness of primary prophylactic ICD implantation. The study also aims for
improved risk stratification and patient selection using clinical risk markers
in general, and advanced ECG risk markers in particular.
| stat |
Minimal sample size in balanced ANOVA models of crossed, nested and
mixed classifications | We consider balanced one-, two- and three-way ANOVA models to test the
hypothesis that the fixed factor A has no effect. The other factors are fixed
or random. We determine the noncentrality parameter for the exact F-test,
describe its minimal value by a sharp lower bound, and thus we can guarantee
the worst case power for the F-test. These results allow us to compute the
minimal sample size. We also provide a structural result for the minimum sample
size, proving a conjecture on the optimal experimental design.
| stat |
Evaluating distributional regression strategies for modelling
self-reported sexual age-mixing | The age dynamics of sexual partnership formation determine patterns of
sexually transmitted disease transmission and have long been a focus of
researchers studying human immunodeficiency virus. Data on self-reported sexual
partner age distributions are available from a variety of sources. We sought to
explore statistical models that accurately predict the distribution of sexual
partner ages over age and sex. We identified which probability distributions
and outcome specifications best captured variation in partner age and
quantified the benefits of modelling these data using distributional
regression. We found that distributional regression with a sinh-arcsinh
distribution replicated observed partner age distributions most accurately
across three geographically diverse data sets. This framework can be extended
with well-known hierarchical modelling tools and can help improve estimates of
sexual age-mixing dynamics.
| stat |
Predicting the Generalization Gap in Deep Networks with Margin
Distributions | As shown in recent research, deep neural networks can perfectly fit randomly
labeled data, but with very poor accuracy on held out data. This phenomenon
indicates that loss functions such as cross-entropy are not a reliable
indicator of generalization. This leads to the crucial question of how
generalization gap should be predicted from the training data and network
parameters. In this paper, we propose such a measure, and conduct extensive
empirical studies on how well it can predict the generalization gap. Our
measure is based on the concept of margin distribution, which are the distances
of training points to the decision boundary. We find that it is necessary to
use margin distributions at multiple layers of a deep network. On the CIFAR-10
and the CIFAR-100 datasets, our proposed measure correlates very strongly with
the generalization gap. In addition, we find the following other factors to be
of importance: normalizing margin values for scale independence, using
characterizations of margin distribution rather than just the margin (closest
distance to decision boundary), and working in log space instead of linear
space (effectively using a product of margins rather than a sum). Our measure
can be easily applied to feedforward deep networks with any architecture and
may point towards new training loss functions that could enable better
generalization.
| stat |
Component-wise approximate Bayesian computation via Gibbs-like steps | Approximate Bayesian computation methods are useful for generative models
with intractable likelihoods. These methods are however sensitive to the
dimension of the parameter space, requiring exponentially increasing resources
as this dimension grows. To tackle this difficulty, we explore a Gibbs version
of the ABC approach that runs component-wise approximate Bayesian computation
steps aimed at the corresponding conditional posterior distributions, and based
on summary statistics of reduced dimensions. While lacking the standard
justifications for the Gibbs sampler, the resulting Markov chain is shown to
converge in distribution under some partial independence conditions. The
associated stationary distribution can further be shown to be close to the true
posterior distribution and some hierarchical versions of the proposed mechanism
enjoy a closed form limiting distribution. Experiments also demonstrate the
gain in efficiency brought by the Gibbs version over the standard solution.
| stat |
An Optimal Private Stochastic-MAB Algorithm Based on an Optimal Private
Stopping Rule | We present a provably optimal differentially private algorithm for the
stochastic multi-arm bandit problem, as opposed to the private analogue of the
UCB-algorithm [Mishra and Thakurta, 2015; Tossou and Dimitrakakis, 2016] which
doesn't meet the recently discovered lower-bound of $\Omega
\left(\frac{K\log(T)}{\epsilon} \right)$ [Shariff and Sheffet, 2018]. Our
construction is based on a different algorithm, Successive Elimination
[Even-Dar et al. 2002], that repeatedly pulls all remaining arms until an arm
is found to be suboptimal and is then eliminated. In order to devise a private
analogue of Successive Elimination we visit the problem of private stopping
rule, that takes as input a stream of i.i.d samples from an unknown
distribution and returns a multiplicative $(1 \pm \alpha)$-approximation of the
distribution's mean, and prove the optimality of our private stopping rule. We
then present the private Successive Elimination algorithm which meets both the
non-private lower bound [Lai and Robbins, 1985] and the above-mentioned private
lower bound. We also compare empirically the performance of our algorithm with
the private UCB algorithm.
| stat |
Understanding and monitoring the evolution of the Covid-19 epidemic from
medical emergency calls: the example of the Paris area | We portray the evolution of the Covid-19 epidemic during the crisis of
March-April 2020 in the Paris area, by analyzing the medical emergency calls
received by the EMS of the four central departments of this area (Centre 15 of
SAMU 75, 92, 93 and 94). Our study reveals strong dissimilarities between these
departments. We show that the logarithm of each epidemic observable can be
approximated by a piecewise linear function of time. This allows us to
distinguish the different phases of the epidemic, and to identify the delay
between sanitary measures and their influence on the load of EMS. This also
leads to an algorithm, allowing one to detect epidemic resurgences. We rely on
a transport PDE epidemiological model, and we use methods from Perron-Frobenius
theory and tropical geometry.
| stat |
Detecting Anomalous Time Series by GAMLSS-Akaike-Weights-Scoring | An extensible statistical framework for detecting anomalous time series
including those with heavy-tailed distributions and non-stationarity in
higher-order moments is introduced based on penalized likelihood distributional
regression. Specifically, generalized additive models for location, scale, and
shape are used to infer sample path representations defined by a parametric
distribution with parameters comprised of basis functions. Akaike weights are
then applied to each model and time series, yielding a probability measure that
can be effectively used to classify and rank anomalous time series. A
mathematical exposition is also given to justify the proposed Akaike weight
scoring under a suitable model embedding as a way to asymptotically identify
anomalous time series. Studies evaluating the methodology on both multiple
simulations and real-world datasets also confirm that high accuracy can be
obtained detecting many different and complex types of shape anomalies. Both
code implementing GAWS for running on a local machine and the datasets
referenced in this paper are available online.
| stat |
Subsets and Splits
No saved queries yet
Save your SQL queries to embed, download, and access them later. Queries will appear here once saved.