title
stringlengths 6
244
| abstract
stringlengths 19
5.09k
| label
stringclasses 10
values |
|---|---|---|
T-EMDE: Sketching-based global similarity for cross-modal retrieval | The key challenge in cross-modal retrieval is to find similarities between
objects represented with different modalities, such as image and text. However,
each modality embeddings stem from non-related feature spaces, which causes the
notorious 'heterogeneity gap'. Currently, many cross-modal systems try to
bridge the gap with self-attention. However, self-attention has been widely
criticized for its quadratic complexity, which prevents many real-life
applications. In response to this, we propose T-EMDE - a neural density
estimator inspired by the recently introduced Efficient Manifold Density
Estimator (EMDE) from the area of recommender systems. EMDE operates on
sketches - representations especially suitable for multimodal operations.
However, EMDE is non-differentiable and ingests precomputed, static embeddings.
With T-EMDE we introduce a trainable version of EMDE which allows full
end-to-end training. In contrast to self-attention, the complexity of our
solution is linear to the number of tokens/segments. As such, T-EMDE is a
drop-in replacement for the self-attention module, with beneficial influence on
both speed and metric performance in cross-modal settings. It facilitates
communication between modalities, as each global text/image representation is
expressed with a standardized sketch histogram which represents the same
manifold structures irrespective of the underlying modality. We evaluate T-EMDE
by introducing it into two recent cross-modal SOTA models and achieving new
state-of-the-art results on multiple datasets and decreasing model latency by
up to 20%.
| stat |
Anomaly detection and classification in traffic flow data from
fluctuations in the flow-density relationship | We describe and validate a novel data-driven approach to the real time
detection and classification of traffic anomalies based on the identification
of atypical fluctuations in the relationship between density and flow. For
aggregated data under stationary conditions, flow and density are related by
the fundamental diagram. However, high resolution data obtained from modern
sensor networks is generally non-stationary and disaggregated. Such data
consequently show significant statistical fluctuations. These fluctuations are
best described using a bivariate probability distribution in the density-flow
plane. By applying kernel density estimation to high-volume data from the UK
National Traffic Information Service (NTIS), we empirically construct these
distributions for London's M25 motorway. Curves in the density-flow plane are
then constructed, analogous to quantiles of univariate distributions. These
curves quantitatively separate atypical fluctuations from typical traffic
states. Although the algorithm identifies anomalies in general rather than
specific events, we find that fluctuations outside the 95\% probability curve
correlate strongly with the spikes in travel time associated with significant
congestion events. Moreover, the size of an excursion from the typical region
provides a simple, real-time measure of the severity of detected anomalies. We
validate the algorithm by benchmarking its ability to identify labelled events
in historical NTIS data against some commonly used methods from the literature.
Detection rate, time-to-detect and false alarm rate are used as metrics and
found to be generally comparable except in situations when the speed
distribution is bi-modal. In such situations, the new algorithm achieves a much
lower false alarm rate without suffering significant degradation on the other
metrics. This method has the additional advantage of being self-calibrating.
| stat |
Considerations Across Three Cultures: Parametric Regressions,
Interpretable Algorithms, and Complex Algorithms | We consider an extension of Leo Breiman's thesis from "Statistical Modeling:
The Two Cultures" to include a bifurcation of algorithmic modeling, focusing on
parametric regressions, interpretable algorithms, and complex (possibly
explainable) algorithms.
| stat |
A Koopman framework for rare event simulation in stochastic differential
equations | We exploit the relationship between the stochastic Koopman operator and the
Kolmogorov backward equation to construct importance sampling schemes for
stochastic differential equations. Specifically, we propose using
eigenfunctions of the stochastic Koopman operator to approximate the Doob
transform for an observable of interest (e.g., associated with a rare event)
which in turn yields an approximation of the corresponding zero-variance
importance sampling estimator. Our approach is broadly applicable and
systematic, treating non-normal systems, non-gradient systems, and systems with
oscillatory dynamics or rank-deficient noise in a common framework. In
nonlinear settings where the stochastic Koopman eigenfunctions cannot be
derived analytically, we use dynamic mode decomposition (DMD) methods to
compute them numerically, but the framework is agnostic to the particular
numerical method employed. Numerical experiments demonstrate that even coarse
approximations of a few eigenfunctions, where the latter are built from
non-rare trajectories, can produce effective importance sampling schemes for
rare events.
| stat |
Population Quasi-Monte Carlo | Monte Carlo methods are widely used for approximating complicated,
multidimensional integrals for Bayesian inference. Population Monte Carlo (PMC)
is an important class of Monte Carlo methods, which utilizes a population of
proposals to generate weighted samples that approximate the target
distribution. The generic PMC framework iterates over three steps: samples are
simulated from a set of proposals, weights are assigned to such samples to
correct for mismatch between the proposal and target distributions, and the
proposals are then adapted via resampling from the weighted samples. When the
target distribution is expensive to evaluate, the PMC has its computational
limitation since the convergence rate is $\mathcal{O}(N^{-1/2})$. To address
this, we propose in this paper a new Population Quasi-Monte Carlo (PQMC)
framework, which integrates Quasi-Monte Carlo ideas within the sampling and
adaptation steps of PMC. A key novelty in PQMC is the idea of importance
support points resampling, a deterministic method for finding an "optimal"
subsample from the weighted proposal samples. Moreover, within the PQMC
framework, we develop an efficient covariance adaptation strategy for
multivariate normal proposals. Lastly, a new set of correction weights is
introduced for the weighted PMC estimator to improve the efficiency from the
standard PMC estimator. We demonstrate the improved empirical convergence of
PQMC over PMC in extensive numerical simulations and a friction drilling
application.
| stat |
Integrative Methods for Post-Selection Inference Under Convex
Constraints | Inference after model selection has been an active research topic in the past
few years, with numerous works offering different approaches to addressing the
perils of the reuse of data. In particular, major progress has been made
recently on large and useful classes of problems by harnessing general theory
of hypothesis testing in exponential families, but these methods have their
limitations. Perhaps most immediate is the gap between theory and practice:
implementing the exact theoretical prescription in realistic situations---for
example, when new data arrives and inference needs to be adjusted
accordingly---may be a prohibitive task. In this paper we propose a Bayesian
framework for carrying out inference after model selection in the linear model.
Our framework is very flexible in the sense that it naturally accommodates
different models for the data, instead of requiring a case-by-case treatment.
At the core of our methods is a new approximation to the exact likelihood
conditional on selection, the latter being generally intractable. We prove
that, under appropriate conditions, our approximation is asymptotically
consistent with the exact truncated likelihood. The advantages of our methods
in practical data analysis are demonstrated in simulations and in application
to HIV drug-resistance data.
| stat |
Protocol for a Study of the Effect of Surface Mining in Central
Appalachia on Adverse Birth Outcomes | Surface mining has become a major method of coal mining in Central Appalachia
alongside the traditional underground mining. Concerns have been raised about
the health effects of this surface mining, particularly mountaintop removal
mining where coal is mined upon steep mountaintops by removing the mountaintop
through clearcutting forests and explosives. We have designed a matched
observational study to assess the effects of surface mining in Central
Appalachia on adverse birth outcomes. This protocol describes for the study the
background and motivation, the sample selection and the analysis plan.
| stat |
PoD-TPI: Probability-of-Decision Toxicity Probability Interval Design to
Accelerate Phase I Trials | Cohort-based enrollment can slow down dose-finding trials since the outcomes
of the previous cohort must be fully evaluated before the next cohort can be
enrolled. This results in frequent suspension of patient enrollment. The issue
is exacerbated in recent immune-oncology trials where toxicity outcomes can
take a long time to observe. We propose a novel phase I design, the
probability-of-decision toxicity probability interval (PoD-TPI) design, to
accelerate phase I trials. PoD-TPI enables dose assignment in real-time in the
presence of pending toxicity outcomes. With uncertain outcomes, the dose
assignment decisions are treated as a random variable, and we calculate the
posterior distribution of the decisions. The posterior distribution reflects
the variability in the pending outcomes and allows a direct and intuitive
evaluation of the confidence of all possible decisions. Optimal decisions are
calculated based on 0-1 loss, and extra safety rules are constructed to enforce
sufficient protection from exposing patients to risky doses. A new and useful
feature of PoD-TPI is that it allows investigators and regulators to balance
the trade-off between enrollment speed and making risky decisions by tuning a
pair of intuitive design parameters. Through numerical studies, we evaluate the
operating characteristics of PoD-TPI and demonstrate that PoD-TPI shortens
trial duration and maintains trial safety and efficiency compared to existing
time-to-event designs.
| stat |
Particle Filtering Under General Regime Switching | In this paper, we consider a new framework for particle filtering under model
uncertainty that operates beyond the scope of Markovian switching systems.
Specifically, we develop a novel particle filtering algorithm that applies to
general regime switching systems, where the model index is augmented as an
unknown time-varying parameter in the system. The proposed approach does not
require the use of multiple filters and can maintain a diverse set of particles
for each considered model through appropriate choice of the particle filtering
proposal distribution. The flexibility of the proposed approach allows for
long-term dependencies between the models, which enables its use to a wider
variety of real-world applications. We validate the method on a synthetic data
experiment and show that it outperforms state-of-the-art multiple model
particle filtering approaches that require the use of multiple filters.
| stat |
Sparse Feature Selection in Kernel Discriminant Analysis via Optimal
Scoring | We consider the two-group classification problem and propose a kernel
classifier based on the optimal scoring framework. Unlike previous approaches,
we provide theoretical guarantees on the expected risk consistency of the
method. We also allow for feature selection by imposing structured sparsity
using weighted kernels. We propose fully-automated methods for selection of all
tuning parameters, and in particular adapt kernel shrinkage ideas for ridge
parameter selection. Numerical studies demonstrate the superior classification
performance of the proposed approach compared to existing nonparametric
classifiers.
| stat |
Learning with risks based on M-location | In this work, we study a new class of risks defined in terms of the location
and deviation of the loss distribution, generalizing far beyond classical
mean-variance risk functions. The class is easily implemented as a wrapper
around any smooth loss, it admits finite-sample stationarity guarantees for
stochastic gradient methods, it is straightforward to interpret and adjust,
with close links to M-estimators of the loss location, and has a salient effect
on the test loss distribution.
| stat |
Optimal sample size for the Birnbaum-Saunders distribution under a
decision-theoretic approach | The Birnbaum-Saunders distribution has been widely applied in several areas
of science and although several methodologies related to this distribution have
been proposed, the problem of determining the optimal sample size for
estimating its mean has not yet been studied. For this purpose, we propose a
methodology to determine the optimal sample size under a decision-theoretic
approach. In this approach, we consider loss functions for point and interval
inference. Finally, computational tools in the R language were developed to use
in practice.
| stat |
Estimating a novel stochastic model for within-field disease dynamics of
banana bunchy top virus via approximate Bayesian computation | The Banana Bunchy Top Virus (BBTV) is one of the most economically important
vector-borne banana diseases throughout the Asia-Pacific Basin and presents a
significant challenge to the agricultural sector. Current models of BBTV are
largely deterministic, limited by an incomplete understanding of interactions
in complex natural systems, and the appropriate identification of parameters. A
stochastic network-based Susceptible-Infected model has been created which
simulates the spread of BBTV across the subsections of a banana plantation,
parameterising nodal recovery, neighbouring and distant infectivity across
summer and winter. Findings from posterior results achieved through Markov
Chain Monte Carlo approach to approximate Bayesian computation suggest
seasonality in all parameters, which are influenced by correlated changes in
inspection accuracy, temperatures and aphid activity. This paper demonstrates
how the model may be used for monitoring and forecasting of various disease
management strategies to support policy-level decision making.
| stat |
Spatio-temporal Sequence Prediction with Point Processes and
Self-organizing Decision Trees | We study the spatio-temporal prediction problem and introduce a novel
point-process-based prediction algorithm. Spatio-temporal prediction is
extensively studied in Machine Learning literature due to its critical
real-life applications such as crime, earthquake, and social event prediction.
Despite these thorough studies, specific problems inherent to the application
domain are not yet fully explored. Here, we address the non-stationary
spatio-temporal prediction problem on both densely and sparsely distributed
sequences. We introduce a probabilistic approach that partitions the spatial
domain into subregions and models the event arrivals in each region with
interacting point-processes. Our algorithm can jointly learn the spatial
partitioning and the interaction between these regions through a gradient-based
optimization procedure. Finally, we demonstrate the performance of our
algorithm on both simulated data and two real-life datasets. We compare our
approach with baseline and state-of-the-art deep learning-based approaches,
where we achieve significant performance improvements. Moreover, we also show
the effect of using different parameters on the overall performance through
empirical results and explain the procedure for choosing the parameters.
| stat |
High-resolution Probabilistic Precipitation Prediction for use in
Climate Simulations | The accurate prediction of precipitation is important to allow for reliable
warnings of flood or drought risk in a changing climate. However, to make
trust-worthy predictions of precipitation, at a local scale, is one of the most
difficult challenges for today's weather and climate models. This is because
important features, such as individual clouds and high-resolution topography,
cannot be resolved explicitly within simulations due to the significant
computational cost of high-resolution simulations. Climate models are typically
run at $\sim$50-100 km resolution which is insufficient to represent local
precipitation events in satisfying detail. Here, we develop a method to make
probabilistic precipitation predictions based on features that climate models
can resolve well and that is not highly sensitive to the approximations used in
individual models. To predict, we will use a temporal compound Poisson
distribution dependent on the output of climate models at a location. We use
the output of Earth System models at coarse resolution $\sim$50 km as input and
train the statistical models towards precipitation observations over Wales at
$\sim$10 km resolution. A Bayesian inferential scheme is provided so that the
compound-Poisson model can be inferred using a
Gibbs-within-Metropolis-Elliptic-Slice sampling scheme which enables us to
quantify the uncertainty of our predictions. In addition, we use a Gaussian
process regressor on the posterior samples of the model parameters, to infer a
spatially coherent model and hence to produce spatially coherent rainfall
prediction. We illustrate the prediction performance of our model by training
over 5 years of the data up to 31st December 1999 and predicting precipitation
for 20 years afterwards for Cardiff and Wales.
| stat |
Uncovering life-course patterns with causal discovery and survival
analysis | We provide a novel approach and an exploratory study for modelling life event
choices and occurrence from a probabilistic perspective through causal
discovery and survival analysis. Our approach is formulated as a bi-level
problem. In the upper level, we build the life events graph, using causal
discovery tools. In the lower level, for the pairs of life events,
time-to-event modelling through survival analysis is applied to model
time-dependent transition probabilities. Several life events were analysed,
such as getting married, buying a new car, child birth, home relocation and
divorce, together with the socio-demographic attributes for survival modelling,
some of which are age, nationality, number of children, number of cars and home
ownership. The data originates from a survey conducted in Dortmund, Germany,
with the questionnaire containing a series of retrospective questions about
residential and employment biography, travel behaviour and holiday trips, as
well as socio-economic characteristic. Although survival analysis has been used
in the past to analyse life-course data, this is the first time that a bi-level
model has been formulated. The inclusion of a causal discovery algorithm in the
upper-level allows us to first identify causal relationships between
life-course events and then understand the factors that might influence
transition rates between events. This is very different from more classic
choice models where causal relationships are subject to expert interpretations
based on model results.
| stat |
A synthetic likelihood-based Laplace approximation for efficient design
of biological processes | Complex models used to describe biological processes in epidemiology and
ecology often have computationally intractable or expensive likelihoods. This
poses significant challenges in terms of Bayesian inference but more
significantly in the design of experiments. Bayesian designs are found by
maximising the expectation of a utility function over a design space, and
typically this requires sampling from or approximating a large number of
posterior distributions. This renders approaches adopted in inference
computationally infeasible to implement in design. Consequently, optimal design
in such fields has been limited to a small number of dimensions or a restricted
range of utility functions. To overcome such limitations, we propose a
synthetic likelihood-based Laplace approximation for approximating utility
functions for models with intractable likelihoods. As will be seen, the
proposed approximation is flexible in that a wide range of utility functions
can be considered, and remains computationally efficient in high dimensions. To
explore the validity of this approximation, an illustrative example from
epidemiology is considered. Then, our approach is used to design experiments
with a relatively large number of observations in two motivating applications
from epidemiology and ecology.
| stat |
Initialization and Regularization of Factorized Neural Layers | Factorized layers--operations parameterized by products of two or more
matrices--occur in a variety of deep learning contexts, including compressed
model training, certain types of knowledge distillation, and multi-head
self-attention architectures. We study how to initialize and regularize deep
nets containing such layers, examining two simple, understudied schemes,
spectral initialization and Frobenius decay, for improving their performance.
The guiding insight is to design optimization routines for these networks that
are as close as possible to that of their well-tuned, non-decomposed
counterparts; we back this intuition with an analysis of how the initialization
and regularization schemes impact training with gradient descent, drawing on
modern attempts to understand the interplay of weight-decay and
batch-normalization. Empirically, we highlight the benefits of spectral
initialization and Frobenius decay across a variety of settings. In model
compression, we show that they enable low-rank methods to significantly
outperform both unstructured sparsity and tensor methods on the task of
training low-memory residual networks; analogs of the schemes also improve the
performance of tensor decomposition techniques. For knowledge distillation,
Frobenius decay enables a simple, overcomplete baseline that yields a compact
model from over-parameterized training without requiring retraining with or
pruning a teacher network. Finally, we show how both schemes applied to
multi-head attention lead to improved performance on both translation and
unsupervised pre-training.
| stat |
An aberration criterion for conditional models | Conditional models with one pair of conditional and conditioned factors in
Mukerjee et al. (2017) are extended to two pairs in this paper. The extension
includes the parametrization, effect hierarchy, sufficient conditions for
universal optimality, aberration, complementary set theory and the strategy for
finding minimum aberration designs. A catalog of 16-run minimum aberration
designs under conditional models is provided. For five to twelve factors, all
16-run minimum aberration designs under conditional models are also minimum
aberration under traditional models.
| stat |
Revisiting the g-null paradox | The parametric g-formula is an approach to estimating causal effects of
sustained treatment strategies from observational data. An often cited
limitation of the parametric g-formula is the g-null paradox: a phenomenon in
which model misspecification in the parametric g-formula is guaranteed under
the conditions that motivate its use (i.e., when identifiability conditions
hold and measured time-varying confounders are affected by past treatment).
Many users of the parametric g-formula know they must acknowledge the g-null
paradox as a limitation when reporting results but still require clarity on its
meaning and implications. Here we revisit the g-null paradox to clarify its
role in causal inference studies. In doing so, we present analytic examples and
a simulation-based illustration of the bias of parametric g-formula estimates
under the conditions associated with this paradox. Our results highlight the
importance of avoiding overly parsimonious models for the components of the
g-formula when using this method.
| stat |
Estimate Sequences for Variance-Reduced Stochastic Composite
Optimization | In this paper, we propose a unified view of gradient-based algorithms for
stochastic convex composite optimization by extending the concept of estimate
sequence introduced by Nesterov. This point of view covers the stochastic
gradient descent method, variants of the approaches SAGA, SVRG, and has several
advantages: (i) we provide a generic proof of convergence for the
aforementioned methods; (ii) we show that this SVRG variant is adaptive to
strong convexity; (iii) we naturally obtain new algorithms with the same
guarantees; (iv) we derive generic strategies to make these algorithms robust
to stochastic noise, which is useful when data is corrupted by small random
perturbations. Finally, we show that this viewpoint is useful to obtain new
accelerated algorithms in the sense of Nesterov.
| stat |
SEAGLE: A Scalable Exact Algorithm for Large-Scale Set-based GxE Tests
in Biobank Data | The explosion of biobank data offers immediate opportunities for
gene-environment (GxE) interaction studies of complex diseases because of the
large sample sizes and the rich collection in genetic and non-genetic
information. However, the extremely large sample size also introduces new
computational challenges in GxE assessment, especially for set-based GxE
variance component (VC) tests, which is a widely used strategy to boost overall
GxE signals and to evaluate the joint GxE effect of multiple variants from a
biologically meaningful unit (e.g., gene). In this work, focusing on continuous
traits, we present SEAGLE, a Scalable Exact AlGorithm for Large-scale set-based
GxE tests, to permit GxE VC tests for biobank-scale data. SEAGLE employs modern
matrix computations to achieve the same "exact" results as the original GxE VC
tests, and does not impose additional assumptions nor relies on approximations.
SEAGLE can easily accommodate sample sizes in the order of $10^5$, is
implementable on standard laptops, and does not require specialized computing
equipment. We demonstrate the performance of SEAGLE using extensive
simulations. We illustrate its utility by conducting genome-wide gene-based GxE
analysis on the Taiwan Biobank data to explore the interaction of gene and
physical activity status on body mass index.
| stat |
Minkowski distances and standardisation for clustering and
classification of high dimensional data | There are many distance-based methods for classification and clustering, and
for data with a high number of dimensions and a lower number of observations,
processing distances is computationally advantageous compared to the raw data
matrix. Euclidean distances are used as a default for continuous multivariate
data, but there are alternatives. Here the so-called Minkowski distances, $L_1$
(city block)-, $L_2$ (Euclidean)-, $L_3$-, $L_4$-, and maximum distances are
combined with different schemes of standardisation of the variables before
aggregating them. Boxplot transformation is proposed, a new transformation
method for a single variable that standardises the majority of observations but
brings outliers closer to the main bulk of the data. Distances are compared in
simulations for clustering by partitioning around medoids, complete and average
linkage, and classification by nearest neighbours, of data with a low number of
observations but high dimensionality. The $L_1$-distance and the boxplot
transformation show good results.
| stat |
Moving-Resting Process with Measurement Error in Animal Movement
Modeling | Statistical modeling of animal movement is of critical importance. The
continuous trajectory of an animal's movements is only observed at discrete,
often irregularly spaced time points. Most existing models cannot handle the
unequal sampling interval naturally and/or do not allow inactivity periods such
as resting or sleeping. The recently proposed moving-resting (MR) model is a
Brownian motion governed by a telegraph process, which allows periods of
inactivity in one state of the telegraph process. The MR model shows promise in
modeling the movements of predators with long inactive periods such as many
felids, but the lack of accommodation of measurement errors seriously prohibits
its application in practice. Here we incorporate measurement errors in the MR
model and derive basic properties of the model. Inferences are based on a
composite likelihood using the Markov property of the chain composed by every
other observed increments. The performance of the method is validated in finite
sample simulation studies. Application to the movement data of a mountain lion
in Wyoming illustrates the utility of the method.
| stat |
Uncertainty Quantification of Darcy Flow through Porous Media using Deep
Gaussian Process | A computational method based on the non-linear Gaussian process (GP), known
as deep Gaussian processes (deep GPs) for uncertainty quantification &
propagation in modelling of flow through heterogeneous porous media is
presented. The method is also used for reducing dimensionality of model output
and consequently emulating highly complex relationship between hydrogeological
properties and reduced order fluid velocity field in a tractable manner. Deep
GPs are multi-layer hierarchical generalisations of GPs with multiple,
infinitely wide hidden layers that are very efficient models for deep learning
and modelling of high-dimensional complex systems by tackling the complexity
through several hidden layers connected with non-linear mappings. According to
this approach, the hydrogeological data is modelled as the output of a
multivariate GP whose inputs are governed by another GP such that each single
layer is either a standard GP or the Gaussian process latent variable model. A
variational approximation framework is used so that the posterior distribution
of the model outputs associated to given inputs can be analytically
approximated. In contrast to the other dimensionality reduction, methods that
do not provide any information about the dimensionality of each hidden layer,
the proposed method automatically selects the dimensionality of each hidden
layer and it can be used to propagate uncertainty obtained in each layer across
the hierarchy. Using this, dimensionality of the full input space consists of
both geometrical parameters of modelling domain and stochastic hydrogeological
parameters can be simultaneously reduced without the need for any
simplifications generally being assumed for stochastic modelling of subsurface
flow problems. It allows estimation of the flow statistics with greatly reduced
computational efforts compared to other stochastic approaches such as Monte
Carlo method.
| stat |
On High Dimensional Covariate Adjustment for Estimating Causal Effects
in Randomized Trials with Survival Outcomes | We study the estimation of the average causal effect (ACE) on the survival
scale where right-censoring exists and high-dimensional covariate information
is available. We propose new estimators using regularized survival regression
and survival random forests (SRF) to make adjustment with high dimensional
covariates to improve efficiency. We study the behavior of general adjusted
estimator when the adjustments are `risk consistent' and `jackknife
compatible'. The theoretical results provided guarantee that the estimators we
proposed are more efficient than the unadjusted one asymptotically when using
SRF for adjustment. The finite sample behavior of our methods are studied by
simulation, and the results are in agreement with our theoretical results. We
also illustrated our methods via analyzing the real data from transplant
research to identify the relative effectiveness of identical sibling donors
compared to unrelated donors with the adjustment of cytogenetic abnormalities.
| stat |
Confidence intervals in general regression models that utilize uncertain
prior information | We consider a general regression model, without a scale parameter. Our aim is
to construct a confidence interval for a scalar parameter of interest $\theta$
that utilizes the uncertain prior information that a distinct scalar parameter
$\tau$ takes the specified value $t$. This confidence interval should have good
coverage properties. It should also have scaled expected length, where the
scaling is with respect to the usual confidence interval, that (a) is
substantially less than 1 when the prior information is correct, (b) has a
maximum value that is not too large and (c) is close to 1 when the data and
prior information are highly discordant. The asymptotic joint distribution of
the maximum likelihood estimators $\theta$ and $\tau$ is similar to the joint
distributions of these estimators in the particular case of a linear regression
with normally distributed errors having known variance. This similarity is used
to construct a confidence interval with the desired properties by using the
confidence interval, computed using the R package ciuupi, that utilizes the
uncertain prior information in this particular linear regression case. An
important practical application of this confidence interval is to a quantal
bioassay carried out to compare two similar compounds. In this context, the
uncertain prior information is that the hypothesis of "parallelism" holds. We
provide extensive numerical results that illustrate the properties of this
confidence interval in this context.
| stat |
Data-Driven Discovery of Coarse-Grained Equations | Statistical (machine learning) tools for equation discovery require large
amounts of data that are typically computer generated rather than
experimentally observed. Multiscale modeling and stochastic simulations are two
areas where learning on simulated data can lead to such discovery. In both, the
data are generated with a reliable but impractical model, e.g., molecular
dynamics simulations, while a model on the scale of interest is uncertain,
requiring phenomenological constitutive relations and ad-hoc approximations. We
replace the human discovery of such models, which typically involves
spatial/stochastic averaging or coarse-graining, with a machine-learning
strategy based on sparse regression that can be executed in two modes. The
first, direct equation-learning, discovers a differential operator from the
whole dictionary. The second, constrained equation-learning, discovers only
those terms in the differential operator that need to be discovered, i.e.,
learns closure approximations. We illustrate our approach by learning a
deterministic equation that governs the spatiotemporal evolution of the
probability density function of a system state whose dynamics are described by
a nonlinear partial differential equation with random inputs. A series of
examples demonstrates the accuracy, robustness, and limitations of our approach
to equation discovery.
| stat |
WRSE -- a non-parametric weighted-resolution ensemble for predicting
individual survival distributions in the ICU | Dynamic assessment of mortality risk in the intensive care unit (ICU) can be
used to stratify patients, inform about treatment effectiveness or serve as
part of an early-warning system. Static risk scoring systems, such as APACHE or
SAPS, have recently been supplemented with data-driven approaches that track
the dynamic mortality risk over time. Recent works have focused on enhancing
the information delivered to clinicians even further by producing full survival
distributions instead of point predictions or fixed horizon risks. In this
work, we propose a non-parametric ensemble model, Weighted Resolution Survival
Ensemble (WRSE), tailored to estimate such dynamic individual survival
distributions. Inspired by the simplicity and robustness of ensemble methods,
the proposed approach combines a set of binary classifiers spaced according to
a decay function reflecting the relevance of short-term mortality predictions.
Models and baselines are evaluated under weighted calibration and
discrimination metrics for individual survival distributions which closely
reflect the utility of a model in ICU practice. We show competitive results
with state-of-the-art probabilistic models, while greatly reducing training
time by factors of 2-9x.
| stat |
Causal structure based root cause analysis of outliers | We describe a formal approach to identify 'root causes' of outliers observed
in $n$ variables $X_1,\dots,X_n$ in a scenario where the causal relation
between the variables is a known directed acyclic graph (DAG). To this end, we
first introduce a systematic way to define outlier scores. Further, we
introduce the concept of 'conditional outlier score' which measures whether a
value of some variable is unexpected *given the value of its parents* in the
DAG, if one were to assume that the causal structure and the corresponding
conditional distributions are also valid for the anomaly. Finally, we quantify
to what extent the high outlier score of some target variable can be attributed
to outliers of its ancestors. This quantification is defined via Shapley values
from cooperative game theory.
| stat |
Educational Note: Paradoxical Collider Effect in the Analysis of
Non-Communicable Disease Epidemiological Data: a reproducible illustration
and web application | Classical epidemiology has focused on the control of confounding but it is
only recently that epidemiologists have started to focus on the bias produced
by colliders. A collider for a certain pair of variables (e.g., an outcome Y
and an exposure A) is a third variable (C) that is caused by both. In a
directed acyclic graph (DAG), a collider is the variable in the middle of an
inverted fork (i.e., the variable C in A -> C <- Y). Controlling for, or
conditioning an analysis on a collider (i.e., through stratification or
regression) can introduce a spurious association between its causes. This
potentially explains many paradoxical findings in the medical literature, where
established risk factors for a particular outcome appear protective. We use an
example from non-communicable disease epidemiology to contextualize and explain
the effect of conditioning on a collider. We generate a dataset with 1,000
observations and run Monte-Carlo simulations to estimate the effect of 24-hour
dietary sodium intake on systolic blood pressure, controlling for age, which
acts as a confounder, and 24-hour urinary protein excretion, which acts as a
collider. We illustrate how adding a collider to a regression model introduces
bias. Thus, to prevent paradoxical associations, epidemiologists estimating
causal effects should be wary of conditioning on colliders. We provide R-code
in easy-to-read boxes throughout the manuscript and a GitHub repository
(https://github.com/migariane/ColliderApp) for the reader to reproduce our
example. We also provide an educational web application allowing real-time
interaction to visualize the paradoxical effect of conditioning on a collider
http://watzilei.com/shiny/collider/.
| stat |
Stein Variational Gaussian Processes | We show how to use Stein variational gradient descent (SVGD) to carry out
inference in Gaussian process (GP) models with non-Gaussian likelihoods and
large data volumes. Markov chain Monte Carlo (MCMC) is extremely
computationally intensive for these situations, but the parametric assumptions
required for efficient variational inference (VI) result in incorrect inference
when they encounter the multi-modal posterior distributions that are common for
such models. SVGD provides a non-parametric alternative to variational
inference which is substantially faster than MCMC. We prove that for GP models
with Lipschitz gradients the SVGD algorithm monotonically decreases the
Kullback-Leibler divergence from the sampling distribution to the true
posterior. Our method is demonstrated on benchmark problems in both regression
and classification, a multimodal posterior, and an air quality example with
550,134 spatiotemporal observations, showing substantial performance
improvements over MCMC and VI.
| stat |
Analyzing the Fine Structure of Distributions | One aim of data mining is the identification of interesting structures in
data. For better analytical results, the basic properties of an empirical
distribution, such as skewness and eventual clipping, i.e. hard limits in value
ranges, need to be assessed. Of particular interest is the question of whether
the data originate from one process or contain subsets related to different
states of the data producing process. Data visualization tools should deliver a
clear picture of the univariate probability density distribution (PDF) for each
feature. Visualization tools for PDFs typically use kernel density estimates
and include both the classical histogram, as well as the modern tools like
ridgeline plots, bean plots and violin plots. If density estimation parameters
remain in a default setting, conventional methods pose several problems when
visualizing the PDF of uniform, multimodal, skewed distributions and
distributions with clipped data, For that reason, a new visualization tool
called the mirrored density plot (MD plot), which is specifically designed to
discover interesting structures in continuous features, is proposed. The MD
plot does not require adjusting any parameters of density estimation, which is
what may make the use of this plot compelling particularly to non-experts. The
visualization tools in question are evaluated against statistical tests with
regard to typical challenges of explorative distribution analysis. The results
of the evaluation are presented using bimodal Gaussian, skewed distributions
and several features with already published PDFs. In an exploratory data
analysis of 12 features describing quarterly financial statements, when
statistical testing poses a great difficulty, only the MD plots can identify
the structure of their PDFs. In sum, the MD plot outperforms the above
mentioned methods.
| stat |
Penalized Variable Selection in Multi-Parameter Regression Survival
Modelling | Multi-parameter regression (MPR) modelling refers to the approach whereby
covariates are allowed to enter the model through multiple distributional
parameters simultaneously. This is in contrast to the standard approaches where
covariates enter through a single parameter (e.g., a location parameter).
Penalized variable selection has received a significant amount of attention in
recent years: methods such as the least absolute shrinkage and selection
operator (LASSO), smoothly clipped absolute deviation (SCAD), and adaptive
LASSO are used to simultaneously select variables and estimate their regression
coefficients. Therefore, in this paper, we develop penalized multi-parameter
regression methods and investigate their associated performance through
simulation studies and real data; as an example, we consider the Weibull MPR
model.
| stat |
Reinforcement Learning for Molecular Design Guided by Quantum Mechanics | Automating molecular design using deep reinforcement learning (RL) holds the
promise of accelerating the discovery of new chemical compounds. Existing
approaches work with molecular graphs and thus ignore the location of atoms in
space, which restricts them to 1) generating single organic molecules and 2)
heuristic reward functions. To address this, we present a novel RL formulation
for molecular design in Cartesian coordinates, thereby extending the class of
molecules that can be built. Our reward function is directly based on
fundamental physical properties such as the energy, which we approximate via
fast quantum-chemical methods. To enable progress towards de-novo molecular
design, we introduce MolGym, an RL environment comprising several challenging
molecular design tasks along with baselines. In our experiments, we show that
our agent can efficiently learn to solve these tasks from scratch by working in
a translation and rotation invariant state-action space.
| stat |
Pursuing Sources of Heterogeneity in Modeling Clustered Population | Researchers often have to deal with heterogeneous population with mixed
regression relationships, increasingly so in the era of data explosion. In such
problems, when there are many candidate predictors, it is not only of interest
to identify the predictors that are associated with the outcome, but also to
distinguish the true sources of heterogeneity, i.e., to identify the predictors
that have different effects among the clusters and thus are the true
contributors to the formation of the clusters. We clarify the concepts of the
source of heterogeneity that account for potential scale differences of the
clusters and propose a regularized finite mixture effects regression to achieve
heterogeneity pursuit and feature selection simultaneously. As the name
suggests, the problem is formulated under an effects-model parameterization, in
which the cluster labels are missing and the effect of each predictor on the
outcome is decomposed to a common effect term and a set of cluster-specific
terms. A constrained sparse estimation of these effects leads to the
identification of both the variables with common effects and those with
heterogeneous effects. We propose an efficient algorithm and show that our
approach can achieve both estimation and selection consistency. Simulation
studies further demonstrate the effectiveness of our method under various
practical scenarios. Three applications are presented, namely, an imaging
genetics study for linking genetic factors and brain neuroimaging traits in
Alzheimer's disease, a public health study for exploring the association
between suicide risk among adolescents and their school district
characteristics, and a sport analytics study for understanding how the salary
levels of baseball players are associated with their performance and
contractual status.
| stat |
Multi-output Gaussian Process Modulated Poisson Processes for Event
Prediction | Prediction of events such as part replacement and failure events plays a
critical role in reliability engineering. Event stream data are commonly
observed in manufacturing and teleservice systems. Designing predictive models
for individual units based on such event streams is challenging and an
under-explored problem. In this work, we propose a non-parametric prognostic
framework for individualized event prediction based on the inhomogeneous
Poisson processes with a multivariate Gaussian convolution process (MGCP) prior
on the intensity functions. The MGCP prior on the intensity functions of the
inhomogeneous Poisson processes maps data from similar historical units to the
current unit under study which facilitates sharing of information and allows
for analysis of flexible event patterns. To facilitate inference, we derive a
variational inference scheme for learning and estimation of parameters in the
resulting MGCP modulated Poisson process model. Experimental results are shown
on both synthetic data as well as real-world data for fleet based event
prediction.
| stat |
The shortest confidence interval for Poisson mean | The existence of the shortest confidence interval for Poisson mean is shown.
The method of obtaining such an interval is presented as well.
| stat |
Markov-chain-inspired search for MH370 | Markov-chain models are constructed for the probabilistic description of the
drift of marine debris from Malaysian Airlines flight MH370. En route from
Kuala Lumpur to Beijing, the MH370 mysteriously disappeared in the southeastern
Indian Ocean on 8 March 2014, somewhere along the arc of the 7th ping ring
around the Inmarsat-3F1 satellite position when the airplane lost contact. The
models are obtained by discretizing the motion of undrogued satellite-tracked
surface drifting buoys from the global historical data bank. A spectral
analysis, Bayesian estimation, and the computation of most probable paths
between the Inmarsat arc and confirmed airplane debris beaching sites are shown
to constrain the crash site, near 25$^{\circ}$S on the Inmarsat arc.
| stat |
Bayesian Multivariate Probability of Success Using Historical Data with
Strict Control of Family-wise Error Rate | Given the cost and duration of phase III and phase IV clinical trials, the
development of statistical methods for go/no-go decisions is vital. In this
paper, we introduce a Bayesian methodology to compute the probability of
success based on the current data of a treatment regimen for the multivariate
linear model. Our approach utilizes a Bayesian seemingly unrelated regression
model, which allows for multiple endpoints to be modeled jointly even if the
covariates between the endpoints are different. Correlations between endpoints
are explicitly modeled. This Bayesian joint modeling approach unifies single
and multiple testing procedures under a single framework. We develop an
approach to multiple testing that asymptotically guarantees strict family-wise
error rate control, and is more powerful than frequentist approaches to
multiplicity. The method effectively yields those of Ibrahim et al. and
Chuang-Stein as special cases, and, to our knowledge, is the only method that
allows for robust sample size determination for multiple endpoints and/or
hypotheses and the only method that provides strict family-wise type I error
control in the presence of multiplicity.
| stat |
Modelling extreme claims via composite models and threshold selection
methods | The existence of large and extreme claims of a non-life insurance portfolio
influences the ability of (re)insurers to estimate the reserve. The excess
over-threshold method provides a way to capture and model the typical behaviour
of insurance claim data. This paper discusses several composite models with
commonly used bulk distributions, combined with a 2-parameter Pareto
distribution above the threshold. We have explored how several threshold
selection methods perform when estimating the reserve as well as the effect of
the choice of bulk distribution, with varying sample size and tail properties.
To investigate this, a simulation study has been performed. Our study shows
that when data are sufficient, the square root rule has the overall best
performance in terms of the quality of the reserve estimate. The second best is
the exponentiality test, especially when the right tail of the data is extreme.
As the sample size becomes small, the simultaneous estimation has the best
performance. Further, the influence of the choice of bulk distribution seems to
be rather large, especially when the distribution is heavy-tailed. Moreover, it
shows that the empirical estimate of $p_{\leq b}$, the probability that a claim
is below the threshold, is more robust than the theoretical one.
| stat |
Adaptive Realized Hyperbolic GARCH Process: Stability and Estimation | In this paper, we propose an Adaptive Realized Hyperbolic GARCH (A-Realized
HYGARCH) process to model the long memory of high-frequency time series with
possible structural breaks. The structural change is modeled by allowing the
intercept to follow the smooth and flexible function form introduced by Gallant
(1984). In addition, stability conditions of the process are investigated. A
Monte Carlo study is investigated in order to illustrate the performance of the
A-Realized HYGARCH process compared to the Realized HYGARCH with or without
structural change.
| stat |
Uncovering Bias in Order Assignment | Many real life situations require a set of items to be repeatedly placed in a
random sequence. In such circumstances, it is often desirable to test whether
such randomization indeed obtains, yet this problem has received very limited
attention in the literature. This paper articulates the key features of this
problem and presents three "untargeted" tests that require no a priori
information from the analyst. These methods are used to analyze the order in
which lottery numbers are drawn in Powerball, the order in which contestants
perform on American Idol, and the order of candidates on primary election
ballots in Texas and West Virginia. In this last application, multiple
deviations from full randomization are detected, with potentially serious
political and legal consequences. The form these deviations take varies,
depending on institutional factors, which sometimes necessitates the use of
tests that exchange power for increased robustness.
| stat |
Out-of-Distribution Generalization with Maximal Invariant Predictor | Out-of-Distribution (OOD) generalization problem is a problem of seeking the
predictor function whose performance in the worst environments is optimal. This
paper makes two contributions to OOD problem. We first use the basic results of
probability to prove maximal Invariant Predictor(MIP) condition, a theoretical
result that can be used to identify the OOD optimal solution. We then use our
MIP to derive inner-environmental Gradient Alignment(IGA) algorithm that can be
used to help seek the OOD optimal predictor. Previous studies that have
investigated the theoretical aspect of the OOD-problem use strong structural
assumptions such as causal DAG. However, in cases involving image datasets, for
example, the identification of hidden structural relations is itself a
difficult problem. Our theoretical results are different from those of many
previous studies in that it can be applied to cases in which the underlying
structure of a dataset is difficult to analyze. We present an extensive
comparison of previous theoretical approaches to the OODproblems based on the
assumptions they make. We also present an extension of the colored-MNIST that
can more accurately represent the pathological OOD situation than the original
version, and demonstrate the superiority of IGA over previous methods on both
the original and the extended version of Colored-MNIST.
| stat |
Spatiotemporal effects of the causal factors on COVID-19 incidences in
the contiguous United States | Since December 2019, the world has been witnessing the gigantic effect of an
unprecedented global pandemic called Severe Acute Respiratory Syndrome
Coronavirus (SARS-CoV-2) - COVID-19. So far, 38,619,674 confirmed cases and
1,093,522 confirmed deaths due to COVID-19 have been reported. In the United
States (US), the cases and deaths are recorded as 7,833,851 and 215,199.
Several timely researches have discussed the local and global effects of the
confounding factors on COVID-19 casualties in the US. However, most of these
studies considered little about the time varying associations between and among
these factors, which are crucial for understanding the outbreak of the present
pandemic. Therefore, this study adopts various relevant approaches, including
local and global spatial regression models and machine learning to explore the
causal effects of the confounding factors on COVID-19 counts in the contiguous
US. Totally five spatial regression models, spatial lag model (SLM), ordinary
least square (OLS), spatial error model (SEM), geographically weighted
regression (GWR) and multiscale geographically weighted regression (MGWR), are
performed at the county scale to take into account the scale effects on
modelling. For COVID-19 cases, ethnicity, crime, and income factors are found
to be the strongest covariates and explain the maximum model variances. For
COVID-19 deaths, both (domestic and international) migration and income factors
play a crucial role in explaining spatial differences of COVID-19 death counts
across counties. The local coefficient of determination (R2) values derived
from the GWR and MGWR models are found very high over the
Wisconsin-Indiana-Michigan (the Great Lake) region, as well as several parts of
Texas, California, Mississippi and Arkansas.
| stat |
Mixed Effects Envelope Models | When multiple measures are collected repeatedly over time, redundancy
typically exists among responses. The envelope method was recently proposed to
reduce the dimension of responses without loss of information in regression
with multivariate responses. It can gain substantial efficiency over the
standard least squares estimator. In this paper, we generalize the envelope
method to mixed effects models for longitudinal data with possibly unbalanced
design and time-varying predictors. We show that our model provides more
efficient estimators than the standard estimators in mixed effects models.
Improved accuracy and efficiency of the proposed method over the standard mixed
effects model estimator are observed in both the simulations and the Action to
Control Cardiovascular Risk in Diabetes (ACCORD) study.
| stat |
Supervised Convex Clustering | Clustering has long been a popular unsupervised learning approach to identify
groups of similar objects and discover patterns from unlabeled data in many
applications. Yet, coming up with meaningful interpretations of the estimated
clusters has often been challenging precisely due to its unsupervised nature.
Meanwhile, in many real-world scenarios, there are some noisy supervising
auxiliary variables, for instance, subjective diagnostic opinions, that are
related to the observed heterogeneity of the unlabeled data. By leveraging
information from both supervising auxiliary variables and unlabeled data, we
seek to uncover more scientifically interpretable group structures that may be
hidden by completely unsupervised analyses. In this work, we propose and
develop a new statistical pattern discovery method named Supervised Convex
Clustering (SCC) that borrows strength from both information sources and guides
towards finding more interpretable patterns via a joint convex fusion penalty.
We develop several extensions of SCC to integrate different types of
supervising auxiliary variables, to adjust for additional covariates, and to
find biclusters. We demonstrate the practical advantages of SCC through
simulations and a case study on Alzheimer's Disease genomics. Specifically, we
discover new candidate genes as well as new subtypes of Alzheimer's Disease
that can potentially lead to better understanding of the underlying genetic
mechanisms responsible for the observed heterogeneity of cognitive decline in
older adults.
| stat |
Cold Posteriors and Aleatoric Uncertainty | Recent work has observed that one can outperform exact inference in Bayesian
neural networks by tuning the "temperature" of the posterior on a validation
set (the "cold posterior" effect). To help interpret this phenomenon, we argue
that commonly used priors in Bayesian neural networks can significantly
overestimate the aleatoric uncertainty in the labels on many classification
datasets. This problem is particularly pronounced in academic benchmarks like
MNIST or CIFAR, for which the quality of the labels is high. For the special
case of Gaussian process regression, any positive temperature corresponds to a
valid posterior under a modified prior, and tuning this temperature is directly
analogous to empirical Bayes. On classification tasks, there is no direct
equivalence between modifying the prior and tuning the temperature, however
reducing the temperature can lead to models which better reflect our belief
that one gains little information by relabeling existing examples in the
training set. Therefore although cold posteriors do not always correspond to an
exact inference procedure, we believe they may often better reflect our true
prior beliefs.
| stat |
The Role of Mutual Information in Variational Classifiers | Overfitting data is a well-known phenomenon related with the generation of a
model that mimics too closely (or exactly) a particular instance of data, and
may therefore fail to predict future observations reliably. In practice, this
behaviour is controlled by various--sometimes heuristics--regularization
techniques, which are motivated by developing upper bounds to the
generalization error. In this work, we study the generalization error of
classifiers relying on stochastic encodings trained on the cross-entropy loss,
which is often used in deep learning for classification problems. We derive
bounds to the generalization error showing that there exists a regime where the
generalization error is bounded by the mutual information between input
features and the corresponding representations in the latent space, which are
randomly generated according to the encoding distribution. Our bounds provide
an information-theoretic understanding of generalization in the so-called class
of variational classifiers, which are regularized by a Kullback-Leibler (KL)
divergence term. These results give theoretical grounds for the highly popular
KL term in variational inference methods that was already recognized to act
effectively as a regularization penalty. We further observe connections with
well studied notions such as Variational Autoencoders, Information Dropout,
Information Bottleneck and Boltzmann Machines. Finally, we perform numerical
experiments on MNIST and CIFAR datasets and show that mutual information is
indeed highly representative of the behaviour of the generalization error.
| stat |
Lockdown effects in US states: an artificial counterfactual approach | We adopt an artificial counterfactual approach to assess the impact of
lockdowns on the short-run evolution of the number of cases and deaths in some
US states. To do so, we explore the different timing in which US states adopted
lockdown policies, and divide them among treated and control groups. For each
treated state, we construct an artificial counterfactual. On average, and in
the very short-run, the counterfactual accumulated number of cases would be two
times larger if lockdown policies were not implemented.
| stat |
Geometric and statistical techniques for projective mapping of chocolate
chip cookies with a large number of consumers | The so-called rapid sensory methods have proved to be useful for the sensory
study of foods by different types of panels, from trained assessors to
unexperienced consumers. Data from these methods have been traditionally
analyzed using statistical techniques, with some recent works proposing the use
of geometric techniques and graph theory. The present work aims to deepen this
line of research introducing a new method, mixing tools from statistics and
graph theory, for the analysis of data from Projective Mapping. In addition, a
large number of n=349 unexperienced consumers is considered for the first time
in Projective Mapping, evaluating nine commercial chocolate chips cookies which
include a blind duplicate of a multinational best-selling brand and seven
private labels. The data obtained are processed using the standard statistical
technique Multiple Factor Analysis (MFA), the recently appeared geometric
method SensoGraph using Gabriel clustering, and the novel variant introduced
here which is based on the pairwise distances between samples. All methods
provide the same groups of samples, with the blind duplicates appearing close
together. Finally, the stability of the results is studied using bootstrapping
and the RV and Mantel coefficients. The results suggest that, even for
unexperienced consumers, highly stable results can be achieved for MFA and
SensoGraph when considering a large enough number of assessors, around 200 for
the consensus map of MFA or the global similarity matrix of SensoGraph.
| stat |
Micro-level network dynamics of scientific collaboration and impact:
relational hyperevent models for the analysis of coauthor networks | We discuss a recently proposed family of statistical network models -
relational hyperevent models (RHEM) - for analyzing team selection and team
performance in scientific coauthor networks. The underlying rationale for using
RHEM in studies of coauthor networks is that scientific collaboration is
intrinsically polyadic, that is, it typically involves teams of any size.
Consequently, RHEM specify publication rates associated with hyperedges
representing groups of scientists of any size. Going beyond previous work on
RHEM for meeting data, we adapt this model family to settings in which
relational hyperevents have a dedicated outcome, such as a scientific paper
with a measurable impact (e.g., the received number of citations). Relational
outcome can on the one hand be used to specify additional explanatory variables
in RHEM since the probability of coauthoring may be influenced, for instance,
by prior (shared) success of scientists. On the other hand relational outcome
can also serve as a response variable in models seeking to explain the
performance of scientific teams. To tackle the latter we propose relational
hyperevent outcome models (RHOM) that are closely related with RHEM to the
point that both model families can specify the likelihood of scientific
collaboration - and the expected performance, respectively - with the same set
of explanatory variables allowing to assess, for instance, whether variables
leading to increased collaboration also tend to increase scientific impact. For
illustration, we apply RHEM to empirical coauthor networks comprising more than
350,000 published papers by scientists working in three scientific disciplines.
| stat |
Stochastic Proximal Langevin Algorithm: Potential Splitting and
Nonasymptotic Rates | We propose a new algorithm---Stochastic Proximal Langevin Algorithm
(SPLA)---for sampling from a log concave distribution. Our method is a
generalization of the Langevin algorithm to potentials expressed as the sum of
one stochastic smooth term and multiple stochastic nonsmooth terms. In each
iteration, our splitting technique only requires access to a stochastic
gradient of the smooth term and a stochastic proximal operator for each of the
nonsmooth terms. We establish nonasymptotic sublinear and linear convergence
rates under convexity and strong convexity of the smooth term, respectively,
expressed in terms of the KL divergence and Wasserstein distance. We illustrate
the efficiency of our sampling technique through numerical simulations on a
Bayesian learning task.
| stat |
Digit analysis for Covid-19 reported data | The coronavirus which appeared in December 2019 in Wuhan has spread out
worldwide and caused the death of more than 280,000 people (as of May, 11
2020). Since February 2020, doubts were raised about the numbers of confirmed
cases and deaths reported by the Chinese government. In this paper, we examine
data available from China at the city and provincial levels and we compare them
with Canadian provincial data, US state data and French regional data. We
consider cumulative and daily numbers of confirmed cases and deaths and examine
these numbers through the lens of their first two digits and in particular we
measure departures of these first two digits to the Newcomb-Benford
distribution, often used to detect frauds. Our finding is that there is no
evidence that cumulative and daily numbers of confirmed cases and deaths for
all these countries have different first or second digit distributions. We also
show that the Newcomb-Benford distribution cannot be rejected for these data.
| stat |
Classifying Sequences of Extreme Length with Constant Memory Applied to
Malware Detection | Recent works within machine learning have been tackling inputs of
ever-increasing size, with cybersecurity presenting sequence classification
problems of particularly extreme lengths. In the case of Windows executable
malware detection, inputs may exceed $100$ MB, which corresponds to a time
series with $T=100,000,000$ steps. To date, the closest approach to handling
such a task is MalConv, a convolutional neural network capable of processing up
to $T=2,000,000$ steps. The $\mathcal{O}(T)$ memory of CNNs has prevented
further application of CNNs to malware. In this work, we develop a new approach
to temporal max pooling that makes the required memory invariant to the
sequence length $T$. This makes MalConv $116\times$ more memory efficient, and
up to $25.8\times$ faster to train on its original dataset, while removing the
input length restrictions to MalConv. We re-invest these gains into improving
the MalConv architecture by developing a new Global Channel Gating design,
giving us an attention mechanism capable of learning feature interactions
across 100 million time steps in an efficient manner, a capability lacked by
the original MalConv CNN. Our implementation can be found at
https://github.com/NeuromorphicComputationResearchProgram/MalConv2
| stat |
Depth with Nonlinearity Creates No Bad Local Minima in ResNets | In this paper, we prove that depth with nonlinearity creates no bad local
minima in a type of arbitrarily deep ResNets with arbitrary nonlinear
activation functions, in the sense that the values of all local minima are no
worse than the global minimum value of corresponding classical machine-learning
models, and are guaranteed to further improve via residual representations. As
a result, this paper provides an affirmative answer to an open question stated
in a paper in the conference on Neural Information Processing Systems 2018.
This paper advances the optimization theory of deep learning only for ResNets
and not for other network architectures.
| stat |
A mathematical framework for raw counts of single-cell RNA-seq data
analysis | Single-cell RNA-seq data are challenging because of the sparseness of the
read counts, the tiny expression of many relevant genes, and the variability in
the efficiency of RNA extraction for different cells. We consider a simple
probabilistic model for read counts, based on a negative binomial distribution
for each gene, modified by a cell-dependent coefficient interpreted as an
extraction efficiency. We provide two alternative fast methods to estimate the
model parameters, together with the probability that a cell results in zero
read counts for a gene. This allows to measure genes co-expression and
differential expression in a novel way.
| stat |
Random Norming Aids Analysis of Non-linear Regression Models with
Sequential Informative Dose Selection | A two-stage adaptive optimal design is an attractive option for increasing
the efficiency of clinical trials. In these designs, based on interim data, the
locally optimal dose is chosen for further exploration, which induces
dependencies between data from the two stages. When the maximum likelihood
estimator (MLE) is used under nonlinear regression models with independent
normal errors in a pilot study where the first stage sample size is fixed, and
the second stage sample size is large, the Fisher information fails to
normalize the estimator adequately asymptotically, because of dependencies. In
this situation, we present three alternative random information measures and
show that they provide better normalization of the MLE asymptotically. The
performance of random information measures is investigated in simulation
studies, and the results suggest that the observed information performs best
when the sample size is small.
| stat |
Distributed Learning of Finite Gaussian Mixtures | Advances in information technology have led to extremely large datasets that
are often kept in different storage centers. Existing statistical methods must
be adapted to overcome the resulting computational obstacles while retaining
statistical validity and efficiency. Split-and-conquer approaches have been
applied in many areas, including quantile processes, regression analysis,
principal eigenspaces, and exponential families. We study split-and-conquer
approaches for the distributed learning of finite Gaussian mixtures. We
recommend a reduction strategy and develop an effective MM algorithm. The new
estimator is shown to be consistent and retains root-n consistency under some
general conditions. Experiments based on simulated and real-world data show
that the proposed split-and-conquer approach has comparable statistical
performance with the global estimator based on the full dataset, if the latter
is feasible. It can even slightly outperform the global estimator if the model
assumption does not match the real-world data. It also has better statistical
and computational performance than some existing methods.
| stat |
Multi-Class Gaussian Process Classification Made Conjugate: Efficient
Inference via Data Augmentation | We propose a new scalable multi-class Gaussian process classification
approach building on a novel modified softmax likelihood function. The new
likelihood has two benefits: it leads to well-calibrated uncertainty estimates
and allows for an efficient latent variable augmentation. The augmented model
has the advantage that it is conditionally conjugate leading to a fast
variational inference method via block coordinate ascent updates. Previous
approaches suffered from a trade-off between uncertainty calibration and speed.
Our experiments show that our method leads to well-calibrated uncertainty
estimates and competitive predictive performance while being up to two orders
faster than the state of the art.
| stat |
Mechanism Design with Bandit Feedback | We study a multi-round welfare-maximising mechanism design problem in
instances where agents do not know their values. On each round, a mechanism
assigns an allocation each to a set of agents and charges them a price; then
the agents provide (stochastic) feedback to the mechanism for the allocation
they received. This is motivated by applications in cloud markets and online
advertising where an agent may know her value for an allocation only after
experiencing it. Therefore, the mechanism needs to explore different
allocations for each agent, while simultaneously attempting to find the
socially optimal set of allocations. Our focus is on truthful and individually
rational mechanisms which imitate the classical VCG mechanism in the long run.
To that end, we define three notions of regret for the welfare, the individual
utilities of each agent and that of the mechanism. We show that these three
terms are interdependent via an $\Omega(T^{\frac{2}{3}})$ lower bound for the
maximum of these three terms after $T$ rounds of allocations, and describe a
family of anytime algorithms which achieve this rate. Our framework provides
flexibility to control the pricing scheme so as to trade-off between the agent
and seller regrets, and additionally to control the degree of truthfulness and
individual rationality.
| stat |
A bivariate likelihood approach for estimation of a pooled continuous
effect size from a heteroscedastic meta-analysis study | The DerSimonian-Laird (DL) weighted average method has been widely used for
estimation of a pooled effect size from an aggregated data meta-analysis study.
It is mainly criticized for its underestimation of the standard error of the
pooled effect size in the presence of heterogeneous study effect sizes. The
uncertainty in the estimation of the between-study variance is not accounted
for in the calculation of this standard error. Due to this negative property,
many alternative estimation approaches have been proposed in literature. One
approach was developed by Hardy and Thompson (HT), who implemented a profile
likelihood approach instead of the moment-based approach of DL. Others have
further extended the likelihood approach and proposed higher-order likelihood
inferences (e.g., Bartlett-type corrections). Likelihood-based methods better
address the uncertainty in estimating the between-study variance than the DL
method, but all these methods assume that the within-study standard deviations
are known and equal to the observed standard error of the study effect sizes.
Here we treat the observed standard errors as estimators for the within-study
variability and we propose a bivariate likelihood approach that jointly
estimates the pooled effect size, the between-study variance, and the
potentially heteroscedastic within-study variances. We study the performance of
the proposed method by means of simulation, and compare it to DL, HT, and the
higher-order likelihood methods. Our proposed approach appears to be less
sensitive to the number of studies, and less biased in case of
heteroscedasticity.
| stat |
Fair Regression for Health Care Spending | The distribution of health care payments to insurance plans has substantial
consequences for social policy. Risk adjustment formulas predict spending in
health insurance markets in order to provide fair benefits and health care
coverage for all enrollees, regardless of their health status. Unfortunately,
current risk adjustment formulas are known to underpredict spending for
specific groups of enrollees leading to undercompensated payments to health
insurers. This incentivizes insurers to design their plans such that
individuals in undercompensated groups will be less likely to enroll, impacting
access to health care for these groups. To improve risk adjustment formulas for
undercompensated groups, we expand on concepts from the statistics, computer
science, and health economics literature to develop new fair regression methods
for continuous outcomes by building fairness considerations directly into the
objective function. We additionally propose a novel measure of fairness while
asserting that a suite of metrics is necessary in order to evaluate risk
adjustment formulas more fully. Our data application using the IBM MarketScan
Research Databases and simulation studies demonstrate that these new fair
regression methods may lead to massive improvements in group fairness (e.g.,
98%) with only small reductions in overall fit (e.g., 4%).
| stat |
Bayesian Learning from Sequential Data using Gaussian Processes with
Signature Covariances | We develop a Bayesian approach to learning from sequential data by using
Gaussian processes (GPs) with so-called signature kernels as covariance
functions. This allows to make sequences of different length comparable and to
rely on strong theoretical results from stochastic analysis. Signatures capture
sequential structure with tensors that can scale unfavourably in sequence
length and state space dimension. To deal with this, we introduce a sparse
variational approach with inducing tensors. We then combine the resulting GP
with LSTMs and GRUs to build larger models that leverage the strengths of each
of these approaches and benchmark the resulting GPs on multivariate time series
(TS) classification datasets. Code available at
https://github.com/tgcsaba/GPSig.
| stat |
Evaluation of Sustainable Green Materials: Pinecone in Permeable
Adsorptive Barriers for Remediation of Groundwater Contaminated by $Pb^{2+}$
and Methylene Blue | We report herein, the potential of raw pinecone powder (PCP) and pinecone
biochar (PCBC) as alternatives to activated carbon used in Permeable Adsorptive
Barriers (PABs) for the in situ remediation of polluted groundwater. A
constructed lab-scale unconfined aquifer ($38\times30\times17$ $cm$) fitted
with PCP and PCBC PABs ($21\times3\times20$ $cm$), was evaluated for the
removal of $Pb^{2+}$ ions in a continuous flow setup. Results indicate that
after $3600$ minutes, PCP was able to reduce $Pb^{2+}$ ions from a Co=$50$
$mg/L$ to $7.94$ $mg/L$ for the first run and $19.4$ $mg/L$ for a second run,
respectively. Comparatively, PCBC reached $6.5$ $mg/L$ for the first run and
$8.94$ for the second run. It was confirmed that adsorption was best described
by the first-order kinetic model with $R^2$ values above $0.95$. Maximum
adsorption capacity values were found to be $1.00$, $0.63$, $1.08$, and $0.85$
mg/g for each scenario respectively. In addition, nonlinear regression models
of exponential and Gaussian Processes are fit to explain remediation by time
for $Pb^{2+}$ and Methylene Blue. Gaussian Processes are able to better explain
the variation of pollution removal compared to simpler exponential models. When
regressed against true removal percentages all models are able to provide
$R^2>0.99$.
| stat |
The Limitations of Adversarial Training and the Blind-Spot Attack | The adversarial training procedure proposed by Madry et al. (2018) is one of
the most effective methods to defend against adversarial examples in deep
neural networks (DNNs). In our paper, we shed some lights on the practicality
and the hardness of adversarial training by showing that the effectiveness
(robustness on test set) of adversarial training has a strong correlation with
the distance between a test point and the manifold of training data embedded by
the network. Test examples that are relatively far away from this manifold are
more likely to be vulnerable to adversarial attacks. Consequentially, an
adversarial training based defense is susceptible to a new class of attacks,
the "blind-spot attack", where the input images reside in "blind-spots" (low
density regions) of the empirical distribution of training data but is still on
the ground-truth data manifold. For MNIST, we found that these blind-spots can
be easily found by simply scaling and shifting image pixel values. Most
importantly, for large datasets with high dimensional and complex data manifold
(CIFAR, ImageNet, etc), the existence of blind-spots in adversarial training
makes defending on any valid test examples difficult due to the curse of
dimensionality and the scarcity of training data. Additionally, we find that
blind-spots also exist on provable defenses including (Wong & Kolter, 2018) and
(Sinha et al., 2018) because these trainable robustness certificates can only
be practically optimized on a limited set of training data.
| stat |
Deep Learning Models to Predict Pediatric Asthma Emergency Department
Visits | Pediatric asthma is the most prevalent chronic childhood illness, afflicting
about 6.2 million children in the United States. However, asthma could be
better managed by identifying and avoiding triggers, educating about
medications and proper disease management strategies. This research utilizes
deep learning methodologies to predict asthma-related emergency department (ED)
visit within 3 months using Medicaid claims data. We compare prediction results
against traditional statistical classification model - penalized Lasso logistic
regression, which we trained and have deployed since 2015. The results have
indicated that deep learning model Artificial Neural Networks (ANN) slightly
outperforms (with AUC = 0.845) the Lasso logistic regression (with AUC =
0.842). The reason may come from the nonlinear nature of ANN.
| stat |
An Improved Online Penalty Parameter Selection Procedure for
$\ell_1$-Penalized Autoregressive with Exogenous Variables | Many recent developments in the high-dimensional statistical time series
literature have centered around time-dependent applications that can be adapted
to regularized least squares. Of particular interest is the lasso, which both
serves to regularize and provide feature selection. The lasso requires the
specification of a penalty parameter that determines the degree of sparsity to
impose. The most popular penalty parameter selection approaches that respect
time dependence are very computationally intensive and are not appropriate for
modeling certain classes of time series. We propose enhancing a canonical time
series model, the autoregressive model with exogenous variables, with a novel
online penalty parameter selection procedure that takes advantage of the
sequential nature of time series data to improve both computational performance
and forecast accuracy relative to existing methods in both a simulation and
empirical application involving macroeconomic indicators.
| stat |
Emulating computer models with step-discontinuous outputs using Gaussian
processes | In many real-world applications we are interested in approximating costly
functions that are analytically unknown, e.g. complex computer codes. An
emulator provides a fast approximation of such functions relying on a limited
number of evaluations. Gaussian processes (GPs) are commonplace emulators due
to their statistical properties such as the ability to estimate their own
uncertainty. GPs are essentially developed to fit smooth, continuous functions.
However, the assumptions of continuity and smoothness is unwarranted in many
situations. For example, in computer models where bifurcations or tipping
points occur, the outputs can be discontinuous. This work examines the capacity
of GPs for emulating step-discontinuous functions. Several approaches are
proposed for this purpose. Two special covariance functions/kernels are adapted
with the ability to model discontinuities. They are the neural network and
Gibbs kernels whose properties are demonstrated using several examples. Another
approach, which is called warping, is to transform the input space into a new
space where a GP with a standard kernel, such as the Matern family, is able to
predict the function well. The transformation is perform by a parametric map
whose parameters are estimated by maximum likelihood. The results show that the
proposed approaches have superior performance to GPs with standard kernels in
capturing sharp jumps in the true function.
| stat |
Gene-Environment Interaction: A Variable Selection Perspective | Gene-environment interactions have important implications to elucidate the
genetic basis of complex diseases beyond the joint function of multiple genetic
factors and their interactions (or epistasis). In the past, G$\times$E
interactions have been mainly conducted within the framework of genetic
association studies. The high dimensionality of G$\times$E interactions, due to
the complicated form of environmental effects and presence of a large number of
genetic factors including gene expressions and SNPs, has motivated the recent
development of penalized variable selection methods for dissecting G$\times$E
interactions, which has been ignored in majority of published reviews on
genetic interaction studies. In this article, we first survey existing
overviews on both gene-environment and gene-gene interactions. Then, after a
brief introduction on the variable selection methods, we review penalization
and relevant variable selection methods in marginal and joint paradigms
respectively under a variety of conceptual models. Discussions on strengths and
limitations, as well as computational aspects of the variable selection methods
tailored for G$\times$E studies have also been provided.
| stat |
Counting Defiers: Examples from Health Care | I propose a finite sample inference procedure that uses a likelihood function
derived from the randomization process within an experiment to conduct
inference on various quantities that capture heterogeneous intervention
effects. One such quantity is the number of defiers---individuals whose
treatment runs counter to the intervention. Results from the literature make
informative inference on this quantity seem impossible, but they rely on
different assumptions and data. I only require data on the cross-tabulations of
a binary intervention and a binary treatment. Replacing the treatment variable
with a more general outcome variable, I can perform inference on important
quantities analogous to the number of defiers. I apply the procedure to test
safety and efficacy in hypothetical drug trials for which the point estimate of
the average intervention effect implies that at least 40 out of 100 individuals
would be saved. In one trial, I infer with 95% confidence that at least 3
individuals would be killed, which could stop the drug from being approved.
| stat |
Robust multiple comparisons against a control group with application in
toxicology | The Dunnett procedure compares several treatment or dose groups with a
control group, while controlling the familywise error rate. When deviations
from the normal distribution and heterogeneous variances occur, the nominal
$\alpha$ level may be violated, and power may be reduced. Various robust
modifications are discussed, whereby the novel most likely transformation
(MLT)-Dunnett version is recommended as almost always appropriate by means of a
simulation study. The MLT-Dunnett is especially useful because it can jointly
and comparably analyse differently scaled endpoints. Furthermore, a related
multiple endpoints test is proposed using the odds ratio as a common effect
size. With the statistical software R, the method is readily applicable using
the CRAN libraries \verb|multcomp| and \verb|mlt|, real data can be easily
analyzed.
| stat |
Sequential Classification with Empirically Observed Statistics | Motivated by real-world machine learning applications, we consider a
statistical classification task in a sequential setting where test samples
arrive sequentially. In addition, the generating distributions are unknown and
only a set of empirically sampled sequences are available to a decision maker.
The decision maker is tasked to classify a test sequence which is known to be
generated according to either one of the distributions. In particular, for the
binary case, the decision maker wishes to perform the classification task with
minimum number of the test samples, so, at each step, she declares that either
hypothesis 1 is true, hypothesis 2 is true, or she requests for an additional
test sample. We propose a classifier and analyze the type-I and type-II error
probabilities. We demonstrate the significant advantage of our sequential
scheme compared to an existing non-sequential classifier proposed by Gutman.
Finally, we extend our setup and results to the multi-class classification
scenario and again demonstrate that the variable-length nature of the problem
affords significant advantages as one can achieve the same set of exponents as
Gutman's fixed-length setting but without having the rejection option.
| stat |
Bayesian Dynamic Estimation of Mortality Schedules in Small Areas | The determination of the shapes of mortality curves, the estimation and
projection of mortality patterns over time, and the investigation of
differences in mortality patterns across different small underdeveloped
populations have received special attention in recent years. The challenges
involved in this type of problems are the common sparsity and the unstable
behavior of observed death counts in small areas (populations). These features
impose many dificulties in the estimation of reasonable mortality schedules. In
this chapter, we present a discussion about this problem and we introduce the
use of relational Bayesian dynamic models for estimating and smoothing
mortality schedules by age and sex. Preliminary results are presented,
including a comparison with a methodology recently proposed in the literature.
The analyzes are based on simulated data as well as mortality data observed in
some Brazilian municipalities.
| stat |
Heterocedasticity-Adjusted Ranking and Thresholding for Large-Scale
Multiple Testing | Standardization has been a widely adopted practice in multiple testing, for
it takes into account the variability in sampling and makes the test statistics
comparable across different study units. However, despite conventional wisdom
to the contrary, we show that there can be a significant loss in information
from basing hypothesis tests on standardized statistics rather than the full
data. We develop a new class of heteroscedasticity--adjusted ranking and
thresholding (HART) rules that aim to improve existing methods by
simultaneously exploiting commonalities and adjusting heterogeneities among the
study units. The main idea of HART is to bypass standardization by directly
incorporating both the summary statistic and its variance into the testing
procedure. A key message is that the variance structure of the alternative
distribution, which is subsumed under standardized statistics, is highly
informative and can be exploited to achieve higher power. The proposed HART
procedure is shown to be asymptotically valid and optimal for false discovery
rate (FDR) control. Our simulation results demonstrate that HART achieves
substantial power gain over existing methods at the same FDR level. We
illustrate the implementation through a microarray analysis of myeloma.
| stat |
High-Dimensional Causal Discovery Under non-Gaussianity | We consider graphical models based on a recursive system of linear structural
equations. This implies that there is an ordering, $\sigma$, of the variables
such that each observed variable $Y_v$ is a linear function of a variable
specific error term and the other observed variables $Y_u$ with $\sigma(u) <
\sigma (v)$. The causal relationships, i.e., which other variables the linear
functions depend on, can be described using a directed graph. It has been
previously shown that when the variable specific error terms are non-Gaussian,
the exact causal graph, as opposed to a Markov equivalence class, can be
consistently estimated from observational data. We propose an algorithm that
yields consistent estimates of the graph also in high-dimensional settings in
which the number of variables may grow at a faster rate than the number of
observations, but in which the underlying causal structure features suitable
sparsity; specifically, the maximum in-degree of the graph is controlled. Our
theoretical analysis is couched in the setting of log-concave error
distributions.
| stat |
Capturing the learning curves of generic features maps for realistic
data sets with a teacher-student model | Teacher-student models provide a powerful framework in which the typical case
performance of high-dimensional supervised learning tasks can be studied in
closed form. In this setting, labels are assigned to data - often taken to be
Gaussian i.i.d. - by a teacher model, and the goal is to characterise the
typical performance of the student model in recovering the parameters that
generated the labels. In this manuscript we discuss a generalisation of this
setting where the teacher and student can act on different spaces, generated
with fixed, but generic feature maps. This is achieved via the rigorous study
of a high-dimensional Gaussian covariate model. Our contribution is two-fold:
First, we prove a rigorous formula for the asymptotic training loss and
generalisation error achieved by empirical risk minimization for this model.
Second, we present a number of situations where the learning curve of the model
captures the one of a \emph{realistic data set} learned with kernel regression
and classification, with out-of-the-box feature maps such as random projections
or scattering transforms, or with pre-learned ones - such as the features
learned by training multi-layer neural networks. We discuss both the power and
the limitations of the Gaussian teacher-student framework as a typical case
analysis capturing learning curves as encountered in practice on real data
sets.
| stat |
An Inverse Optimization Approach to Measuring Clinical Pathway
Concordance | Clinical pathways outline standardized processes in the delivery of care for
a specific disease. Patient journeys through the healthcare system, though, can
deviate substantially from these pathways. Given the positive benefits of
clinical pathways, it is important to measure the concordance of patient
pathways so that variations in health system performance or bottlenecks in the
delivery of care can be detected, monitored, and acted upon. This paper
proposes the first data-driven inverse optimization approach to measuring
pathway concordance in any problem context. Our specific application considers
clinical pathway concordance for stage III colon cancer. We develop a novel
concordance metric and demonstrate using real patient data from Ontario, Canada
that it has a statistically significant association with survival. Our
methodological approach considers a patient's journey as a walk in a directed
graph, where the costs on the arcs are derived by solving an inverse shortest
path problem. The inverse optimization model uses two sources of information to
find the arc costs: reference pathways developed by a provincial cancer agency
(primary) and data from real-world patient-related activity from patients with
both positive and negative clinical outcomes (secondary). Thus, our inverse
optimization framework extends existing models by including data points of both
varying "primacy" and "alignment". Data primacy is addressed through a
two-stage approach to imputing the cost vector, while data alignment is
addressed by a hybrid objective function that aims to minimize and maximize
suboptimality error for different subsets of input data.
| stat |
Shifted Randomized Singular Value Decomposition | We extend the randomized singular value decomposition (SVD) algorithm
\citep{Halko2011finding} to estimate the SVD of a shifted data matrix without
explicitly constructing the matrix in the memory. With no loss in the accuracy
of the original algorithm, the extended algorithm provides for a more efficient
way of matrix factorization. The algorithm facilitates the low-rank
approximation and principal component analysis (PCA) of off-center data
matrices. When applied to different types of data matrices, our experimental
results confirm the advantages of the extensions made to the original
algorithm.
| stat |
Dependent censoring based on copulas | Consider a survival time T that is subject to random right censoring, and
suppose that T is stochastically dependent on the censoring time C. We are
interested in the marginal distribution of T. This situation is often
encountered in practice. Consider for instance the case where T is the time to
death of a patient suffering from a certain disease. Then, the censoring time C
is for instance the time until the person leaves the study or the time until
he/she dies from another disease. If the reason for leaving the study is
related to the health condition of the patient or if he/she dies from a disease
that has similar risk factors as the disease of interest, then T and C are
likely dependent. In this paper we propose a new model that takes this
dependence into account. The model is based on a parametric copula for the
relationship between T and C, and on parametric marginal distributions for T
and C. Unlike most other papers in the literature, we do not assume that the
parameter defining the copula function is known. We give sufficient conditions
on these parametric copula and marginals under which the bivariate distribution
of (T;C) is identifed. These sufficient conditions are then checked for a wide
range of common copulas and marginal distributions. We also study the
estimation of the model, and carry out extensive simulations and the analysis
of data on pancreas cancer to illustrate the proposed model and estimation
procedure.
| stat |
Causal Inferences in Small Area Estimation | When doing impact evaluation and making causal inferences, it is important to
acknowledge the heterogeneity of the treatment effects for different domains
(geographic, socio-demographic, or socio-economic). If the domain of interest
is small with regards to its sample size (or even zero in some cases), then the
evaluator has entered the small area estimation (SAE) dilemma.
Based on the modification of the Inverse Propensity Weighting estimator and
the traditional small area predictors, the paper proposes a new methodology to
estimate area specific average treatment effects for unplanned domains. By
means of these methods we can also provide a map of policy impacts, that can
help to better target the treatment group(s). We develop analytical Mean
Squared Error (MSE) estimators of the proposed predictors. An extensive
simulation analysis, also based on real data, shows that the proposed
techniques in most cases lead to more efficient estimators.
| stat |
Efficient Computation of Linear Model Treatment Effects in an
Experimentation Platform | Linear models are a core component for statistical software that analyzes
treatment effects. They are used in experimentation platforms where analysis is
automated, as well as scientific studies where analysis is done locally and
manually. However, the implementations of linear models for online platforms
and for local analysis are often different due to differing demands in
scalability. To improve leverage, we developed one library that powers both
modes of analysis, consolidates many forms of treatment effects from linear
models, and optimizes the computational performance of the software. The highly
performant algorithms allow the library to run linear models at Netflix's scale
and reply to live queries for treatment effects from an online service without
using distributed systems in seconds. Wrapping these algorithms in a library
that can be used interactively make large scale linear models accessible for
both online platforms and scientists to use locally.
| stat |
Quantification of predictive uncertainty in hydrological modelling by
harnessing the wisdom of the crowd: Methodology development and investigation
using toy models | We introduce an ensemble learning post-processing methodology for
probabilistic hydrological modelling. This methodology generates numerous point
predictions by applying a single hydrological model, yet with different
parameter values drawn from the respective simulated posterior distribution. We
call these predictions "sister predictions". Each sister prediction extending
in the period of interest is converted into a probabilistic prediction using
information about the hydrological model's errors. This information is obtained
from a preceding period for which observations are available, and is exploited
using a flexible quantile regression model. All probabilistic predictions are
finally combined via simple quantile averaging to produce the output
probabilistic prediction. The idea is inspired by the ensemble learning methods
originating from the machine learning literature. The proposed methodology
offers larger robustness in performance than basic post-processing
methodologies using a single hydrological point prediction. It is also
empirically proven to "harness the wisdom of the crowd" in terms of average
interval score, i.e., the obtained quantile predictions score no worse --
usually better -- than the average score of the combined individual
predictions. This proof is provided within toy examples, which can be used for
gaining insight on how the methodology works and under which conditions it can
optimally convert point hydrological predictions to probabilistic ones. A
large-scale hydrological application is made in a companion paper.
| stat |
Dynamic Quantile Function Models | Motivated by the need for effectively summarising, modelling, and forecasting
the distributional characteristics of intra-daily returns, as well as the
recent work on forecasting histogram-valued time-series in the area of symbolic
data analysis, we develop a time-series model for forecasting
quantile-function-valued (QF-valued) daily summaries for intra-daily returns.
We call this model the dynamic quantile function (DQF) model. Instead of a
histogram, we propose to use a $g$-and-$h$ quantile function to summarise the
distribution of intra-daily returns. We work with a Bayesian formulation of the
DQF model in order to make statistical inference while accounting for parameter
uncertainty; an efficient MCMC algorithm is developed for sampling-based
posterior inference. Using ten international market indices and approximately
2,000 days of out-of-sample data from each market, the performance of the DQF
model compares favourably, in terms of forecasting VaR of intra-daily returns,
against the interval-valued and histogram-valued time-series models.
Additionally, we demonstrate that the QF-valued forecasts can be used to
forecast VaR measures at the daily timescale via a simple quantile regression
model on daily returns (QR-DQF). In certain markets, the resulting QR-DQF model
is able to provide competitive VaR forecasts for daily returns.
| stat |
Designing Transportable Experiments | We consider the problem of designing a randomized experiment on a source
population to estimate the Average Treatment Effect (ATE) on a target
population. We propose a novel approach which explicitly considers the target
when designing the experiment on the source. Under the covariate shift
assumption, we design an unbiased importance-weighted estimator for the target
population's ATE. To reduce the variance of our estimator, we design a
covariate balance condition (Target Balance) between the treatment and control
groups based on the target population. We show that Target Balance achieves a
higher variance reduction asymptotically than methods that do not consider the
target population during the design phase. Our experiments illustrate that
Target Balance reduces the variance even for small sample sizes.
| stat |
Geom-SPIDER-EM: Faster Variance Reduced Stochastic Expectation
Maximization for Nonconvex Finite-Sum Optimization | The Expectation Maximization (EM) algorithm is a key reference for inference
in latent variable models; unfortunately, its computational cost is prohibitive
in the large scale learning setting. In this paper, we propose an extension of
the Stochastic Path-Integrated Differential EstimatoR EM (SPIDER-EM) and derive
complexity bounds for this novel algorithm, designed to solve smooth nonconvex
finite-sum optimization problems. We show that it reaches the same state of the
art complexity bounds as SPIDER-EM; and provide conditions for a linear rate of
convergence. Numerical results support our findings.
| stat |
Hierarchical Indian Buffet Neural Networks for Bayesian Continual
Learning | We place an Indian Buffet process (IBP) prior over the structure of a
Bayesian Neural Network (BNN), thus allowing the complexity of the BNN to
increase and decrease automatically. We further extend this model such that the
prior on the structure of each hidden layer is shared globally across all
layers, using a Hierarchical-IBP (H-IBP). We apply this model to the problem of
resource allocation in Continual Learning (CL) where new tasks occur and the
network requires extra resources. Our model uses online variational inference
with reparameterisation of the Bernoulli and Beta distributions, which
constitute the IBP and H-IBP priors. As we automatically learn the number of
weights in each layer of the BNN, overfitting and underfitting problems are
largely overcome. We show empirically that our approach offers a competitive
edge over existing methods in CL.
| stat |
Applications of Clustering with Mixed Type Data in Life Insurance | Death benefits are generally the largest cash flow item that affects
financial statements of life insurers where some still do not have a systematic
process to track and monitor death claims experience. In this article, we
explore data clustering to examine and understand how actual death claims
differ from expected, an early stage of developing a monitoring system crucial
for risk management. We extend the $k$-prototypes clustering algorithm to draw
inference from a life insurance dataset using only the insured's
characteristics and policy information without regard to known mortality. This
clustering has the feature to efficiently handle categorical, numerical, and
spatial attributes. Using gap statistics, the optimal clusters obtained from
the algorithm are then used to compare actual to expected death claims
experience of the life insurance portfolio. Our empirical data contains
observations, during 2014, of approximately 1.14 million policies with a total
insured amount of over 650 billion dollars. For this portfolio, the algorithm
produced three natural clusters, with each cluster having a lower actual to
expected death claims but with differing variability. The analytical results
provide management a process to identify policyholders' attributes that
dominate significant mortality deviations, and thereby enhance decision making
for taking necessary actions.
| stat |
Equivalence relations and $L^p$ distances between time series | We introduce a general framework for defining equivalence and measuring
distances between time series, and a first concrete method for doing so. We
prove the existence of equivalence relations on the space of time series, such
that the quotient spaces can be equipped with a metrizable topology. We
illustrate algorithmically how to calculate such distances among a collection
of time series, and perform clustering analysis based on these distances. We
apply these insights to analyse the recent bushfires in NSW, Australia. There,
we introduce a new method to analyse time series in a cross-contextual setting.
| stat |
Structural Agnostic Modeling: Adversarial Learning of Causal Graphs | A new causal discovery method, Structural Agnostic Modeling (SAM), is
presented in this paper. Leveraging both conditional independencies and
distributional asymmetries in the data, SAM aims to find the underlying causal
structure from observational data. The approach is based on a game between
different players estimating each variable distribution conditionally to the
others as a neural net, and an adversary aimed at discriminating the overall
joint conditional distribution, and that of the original data. A learning
criterion combining distribution estimation, sparsity and acyclicity
constraints is used to enforce the end-to-end optimization of the graph
structure and parameters through stochastic gradient descent. Besides a
theoretical analysis of the approach in the large sample limit, SAM is
extensively experimentally validated on synthetic and real data.
| stat |
Modeling microbial abundances and dysbiosis with beta-binomial
regression | Using a sample from a population to estimate the proportion of the population
with a certain category label is a broadly important problem. In the context of
microbiome studies, this problem arises when researchers wish to use a sample
from a population of microbes to estimate the population proportion of a
particular taxon, known as the taxon's relative abundance. In this paper, we
propose a beta-binomial model for this task. Like existing models, our model
allows for a taxon's relative abundance to be associated with covariates of
interest. However, unlike existing models, our proposal also allows for the
overdispersion in the taxon's counts to be associated with covariates of
interest. We exploit this model in order to propose tests not only for
differential relative abundance, but also for differential variability. The
latter is particularly valuable in light of speculation that dysbiosis, the
perturbation from a normal microbiome that can occur in certain disease
conditions, may manifest as a loss of stability, or increase in variability, of
the counts associated with each taxon. We demonstrate the performance of our
proposed model using a simulation study and an application to soil microbial
data.
| stat |
A Bayesian nonparametric approach to count-min sketch under power-law
data streams | The count-min sketch (CMS) is a randomized data structure that provides
estimates of tokens' frequencies in a large data stream using a compressed
representation of the data by random hashing. In this paper, we rely on a
recent Bayesian nonparametric (BNP) view on the CMS to develop a novel
learning-augmented CMS under power-law data streams. We assume that tokens in
the stream are drawn from an unknown discrete distribution, which is endowed
with a normalized inverse Gaussian process (NIGP) prior. Then, using
distributional properties of the NIGP, we compute the posterior distribution of
a token's frequency in the stream, given the hashed data, and in turn
corresponding BNP estimates. Applications to synthetic and real data show that
our approach achieves a remarkable performance in the estimation of
low-frequency tokens. This is known to be a desirable feature in the context of
natural language processing, where it is indeed common in the context of the
power-law behaviour of the data.
| stat |
Real-time outlier detection for large datasets by RT-DetMCD | Modern industrial machines can generate gigabytes of data in seconds,
frequently pushing the boundaries of available computing power. Together with
the time criticality of industrial processing this presents a challenging
problem for any data analytics procedure. We focus on the deterministic minimum
covariance determinant method (DetMCD), which detects outliers by fitting a
robust covariance matrix. We construct a much faster version of DetMCD by
replacing its initial estimators by two new methods and incorporating
update-based concentration steps. The computation time is reduced further by
parallel computing, with a novel robust aggregation method to combine the
results from the threads. The speed and accuracy of the proposed real-time
DetMCD method (RT-DetMCD) are illustrated by simulation and a real industrial
application to food sorting.
| stat |
Kernel k-Groups via Hartigan's Method | Energy statistics was proposed by Sz\' ekely in the 80's inspired by Newton's
gravitational potential in classical mechanics and it provides a model-free
hypothesis test for equality of distributions. In its original form, energy
statistics was formulated in Euclidean spaces. More recently, it was
generalized to metric spaces of negative type. In this paper, we consider a
formulation for the clustering problem using a weighted version of energy
statistics in spaces of negative type. We show that this approach leads to a
quadratically constrained quadratic program in the associated kernel space,
establishing connections with graph partitioning problems and kernel methods in
machine learning. To find local solutions of such an optimization problem, we
propose kernel k-groups, which is an extension of Hartigan's method to kernel
spaces. Kernel k-groups is cheaper than spectral clustering and has the same
computational cost as kernel k-means (which is based on Lloyd's heuristic) but
our numerical results show an improved performance, especially in higher
dimensions. Moreover, we verify the efficiency of kernel k-groups in community
detection in sparse stochastic block models which has fascinating applications
in several areas of science.
| stat |
On a simultaneous parameter inference and missing data imputation for
nonstationary autoregressive models | This work addresses the problem of missing data in time-series analysis
focusing on (a) estimation of model parameters in the presence of missing data
and (b) reconstruction of missing data. Standard approaches used to solve these
problems like the maximum likelihood estimation or the Bayesian inference rely
on a priori assumptions like the Gaussian or stationary behavior of missing
data and might lead to biased results where these assumptions are unfulfilled.
In order to go beyond, we extend the Finite Element Methodology (FEM) for
Vector Auto-Regressive models with eXogenous factors and bounded variation of
the model parameters (FEM-VARX) towards handling the missing data problem. The
presented approach estimates the model parameters and reconstructs the missing
data in the considered time series and in the involved exogenous factors,
simultaneously. The resulting computational framework was compared to the
state-of-art methodologies on a set of test-cases and is available as
open-source software.
| stat |
A safety factor approach to designing urban infrastructure for dynamic
conditions | Current approaches to design flood-sensitive infrastructure typically assume
a stationary rainfall distribution and neglect many uncertainties. These
assumptions are inconsistent with observations that suggest intensifying
extreme precipitation events and the uncertainties surrounding projections of
the coupled natural-human systems. Here we show that assuming climate
stationarity and neglecting deep uncertainties can drastically underestimate
flood risks and lead to poor infrastructure design choices. We find that
climate uncertainty dominates the socioeconomic and engineering uncertainties
that impact the hydraulic reliability in stormwater drainage systems. We
quantify the upfront costs needed to achieve higher hydraulic reliability and
robustness against the deep uncertainties surrounding projections of rainfall,
surface runoff characteristics, and infrastructure lifetime. Depending on the
location, we find that adding safety factors of 1.4 to 1.7 to the standard
stormwater pipe design guidance produces robust performance to the considered
deep uncertainties.
| stat |
DPPNet: Approximating Determinantal Point Processes with Deep Networks | Determinantal Point Processes (DPPs) provide an elegant and versatile way to
sample sets of items that balance the point-wise quality with the set-wise
diversity of selected items. For this reason, they have gained prominence in
many machine learning applications that rely on subset selection. However,
sampling from a DPP over a ground set of size $N$ is a costly operation,
requiring in general an $O(N^3)$ preprocessing cost and an $O(Nk^3)$ sampling
cost for subsets of size $k$. We approach this problem by introducing DPPNets:
generative deep models that produce DPP-like samples for arbitrary ground sets.
We develop an inhibitive attention mechanism based on transformer networks that
captures a notion of dissimilarity between feature vectors. We show
theoretically that such an approximation is sensible as it maintains the
guarantees of inhibition or dissimilarity that makes DPPs so powerful and
unique. Empirically, we demonstrate that samples from our model receive high
likelihood under the more expensive DPP alternative.
| stat |
Markov Decision Processes with Dynamic Transition Probabilities: An
Analysis of Shooting Strategies in Basketball | In this paper we model basketball plays as episodes from team-specific
non-stationary Markov decision processes (MDPs) with shot clock dependent
transition probabilities. Bayesian hierarchical models are employed in the
modeling and parametrization of the transition probabilities to borrow strength
across players and through time. To enable computational feasibility, we
combine lineup-specific MDPs into team-average MDPs using a novel transition
weighting scheme. Specifically, we derive the dynamics of the team-average
process such that the expected transition count for an arbitrary state-pair is
equal to the weighted sum of the expected counts of the separate
lineup-specific MDPs.
We then utilize these non-stationary MDPs in the creation of a basketball
play simulator with uncertainty propagated via posterior samples of the model
components. After calibration, we simulate seasons both on-policy and under
altered policies and explore the net changes in efficiency and production under
the alternate policies. Additionally, we discuss the game-theoretic
ramifications of testing alternative decision policies.
| stat |
Composable Effects for Flexible and Accelerated Probabilistic
Programming in NumPyro | NumPyro is a lightweight library that provides an alternate NumPy backend to
the Pyro probabilistic programming language with the same modeling interface,
language primitives and effect handling abstractions. Effect handlers allow
Pyro's modeling API to be extended to NumPyro despite its being built atop a
fundamentally different JAX-based functional backend. In this work, we
demonstrate the power of composing Pyro's effect handlers with the program
transformations that enable hardware acceleration, automatic differentiation,
and vectorization in JAX. In particular, NumPyro provides an iterative
formulation of the No-U-Turn Sampler (NUTS) that can be end-to-end JIT
compiled, yielding an implementation that is much faster than existing
alternatives in both the small and large dataset regimes.
| stat |
Predicting Covid-19 EMS Incidents from Daily Hospitalization Trends | Introduction: The aim of our retrospective study was to quantify the impact
of Covid-19 on the spatiotemporal distribution of Emergency Medical Services
(EMS) demands in Travis County, Austin, Texas and propose a robust model to
forecast Covid-19 EMS incidents in the short term to improve EMS performance.
Methods: We analyzed the number of EMS calls and daily Covid-19 hospitalization
in the Austin-Travis County area between January 1st, 2019 and December 31st,
2020. Change point detection was performed to identify critical dates marking
changes in EMS call distributions and time series regression was applied for
our prediction model. Results: Two critical dates mark the impact of Covid-19
on EMS calls: March 17th, when the daily number of Non-Pandemic EMS incidents
dropped significantly, and May 13th, by which the daily number of EMS calls
climbed back to 75% of pre-Covid-19 demand. New daily Covid-19-hospitalization
alone proves a powerful predictor of pandemic EMS calls, with an $r^2$ value
equal to 0.85. Conclusion: The mean daily number of non-pandemic EMS demands
was significantly less than the period prior to Covid-19 pandemic. The number
of EMS calls for Covid-19 symptoms can be predicted from the daily new
hospitalization of Covid-19 patients. In particular, for every 2.5 cases where
EMS takes a Covid-19 patient to a hospital, 1 person is admitted.
| stat |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.