title
stringlengths 6
244
| abstract
stringlengths 19
5.09k
| label
stringclasses 10
values |
|---|---|---|
One-pass Stochastic Gradient Descent in Overparametrized Two-layer
Neural Networks | There has been a recent surge of interest in understanding the convergence of
gradient descent (GD) and stochastic gradient descent (SGD) in
overparameterized neural networks. Most previous works assume that the training
data is provided a priori in a batch, while less attention has been paid to the
important setting where the training data arrives in a stream. In this paper,
we study the streaming data setup and show that with overparamterization and
random initialization, the prediction error of two-layer neural networks under
one-pass SGD converges in expectation. The convergence rate depends on the
eigen-decomposition of the integral operator associated with the so-called
neural tangent kernel (NTK). A key step of our analysis is to show a random
kernel function converges to the NTK with high probability using the VC
dimension and McDiarmid's inequality.
| stat |
Debiased Inverse-Variance Weighted Estimator in Two-Sample Summary-Data
Mendelian Randomization | Mendelian randomization (MR) has become a popular approach to study the
effect of a modifiable exposure on an outcome by using genetic variants as
instrumental variables. A challenge in MR is that each genetic variant explains
a relatively small proportion of variance in the exposure and there are many
such variants, a setting known as many weak instruments. To this end, we
provide a theoretical characterization of the statistical properties of two
popular estimators in MR, the inverse-variance weighted (IVW) estimator and the
IVW estimator with screened instruments using an independent selection dataset,
under many weak instruments. We then propose a debiased IVW estimator, a simple
modification of the IVW estimator, that is robust to many weak instruments and
doesn't require screening. Additionally, we present two instrument selection
methods to improve the efficiency of the new estimator when a selection dataset
is available. An extension of the debiased IVW estimator to handle balanced
horizontal pleiotropy is also discussed. We conclude by demonstrating our
results in simulated and real datasets.
| stat |
Extending the Heston Model to Forecast Motor Vehicle Collision Rates | We present an alternative approach to the forecasting of motor vehicle
collision rates. We adopt an oft-used tool in mathematical finance, the Heston
Stochastic Volatility model, to forecast the short-term and long-term evolution
of motor vehicle collision rates. We incorporate a number of extensions to the
Heston model to make it fit for modelling motor vehicle collision rates. We
incorporate the temporally-unstable and non-deterministic nature of collision
rate fluctuations, and introduce a parameter to account for periods of
accelerated safety. We also adjust estimates to account for the seasonality of
collision patterns. Using these parameters, we perform a short-term forecast of
collision rates and explore a number of plausible scenarios using long-term
forecasts. The short-term forecast shows a close affinity with realised rates
(95% accuracy). The long-term scenarios suggest that modest targets to reduce
collision rates (1.83% annually) and targets to reduce the fluctuations of
month-to-month collision rates (by half) could have significant benefits for
road safety. The median forecast in this scenario suggests a 50% fall in
collision rates, with 75% of simulations suggesting that an effective change in
collision rates is observed before 2044. The main benefit the model provides is
eschewing the necessity for setting unreasonable safety targets that are often
missed. Instead, the model presents the effects that modest and achievable
targets can have on road safety over the long run, while incorporating random
variability. Examining the parameters that underlie expected collision rates
will aid policymakers in determining the effectiveness of implemented policies.
| stat |
Parallel cross-validation: a scalable fitting method for Gaussian
process models | Gaussian process (GP) models are widely used to analyze spatially referenced
data and to predict values at locations without observations. In contrast to
many algorithmic procedures, GP models are based on a statistical framework,
which enables uncertainty quantification of the model structure and
predictions. Both the evaluation of the likelihood and the prediction involve
solving linear systems. Hence, the computational costs are large and limit the
amount of data that can be handled. While there are many approximation
strategies that lower the computational cost of GP models, they often provide
only sub-optimal support for the parallel computing capabilities of current
(high-performance) computing environments. We aim at bridging this gap with a
parameter estimation and prediction method that is designed to be
parallelizable. More precisely, we divide the spatial domain into overlapping
subsets and use cross-validation (CV) to estimate the covariance parameters in
parallel. We present simulation studies, which assess the accuracy of the
parameter estimates and predictions. Moreover, we show that our implementation
has good weak and strong parallel scaling properties. For illustration, we fit
an exponential covariance model to a scientifically relevant canopy height
dataset with 5 million observations. Using 512 processor cores in parallel
brings the evaluation time of one covariance parameter configuration to less
than 1.5 minutes. The parallel CV method can be easily extended to include
approximate likelihood methods, multivariate and spatio-temporal data, as well
as non-stationary covariance models.
| stat |
Uncertainty-aware Sensitivity Analysis Using R\'enyi Divergences | For nonlinear supervised learning models, assessing the importance of
predictor variables or their interactions is not straightforward because it can
vary in the domain of the variables. Importance can be assessed locally with
sensitivity analysis using general methods that rely on the model's predictions
or their derivatives. In this work, we extend derivative based sensitivity
analysis to a Bayesian setting by differentiating the R\'enyi divergence of a
model's predictive distribution. By utilising the predictive distribution
instead of a point prediction, the model uncertainty is taken into account in a
principled way. Our empirical results on simulated and real data sets
demonstrate accurate and reliable identification of important variables and
interaction effects compared to alternative methods.
| stat |
A Survey of Bayesian Statistical Approaches for Big Data | The modern era is characterised as an era of information or Big Data. This
has motivated a huge literature on new methods for extracting information and
insights from these data. A natural question is how these approaches differ
from those that were available prior to the advent of Big Data. We present a
review of published studies that present Bayesian statistical approaches
specifically for Big Data and discuss the reported and perceived benefits of
these approaches. We conclude by addressing the question of whether focusing
only on improving computational algorithms and infrastructure will be enough to
face the challenges of Big Data.
| stat |
Value of information from vibration-based structural health monitoring
extracted via Bayesian model updating | Quantifying the value of the information extracted from a structural health
monitoring (SHM) system is an important step towards convincing decision makers
to implement these systems. We quantify this value by adaptation of the
Bayesian decision analysis framework. In contrast to previous works, we model
in detail the entire process of data generation to processing, model updating
and reliability calculation, and investigate it on a deteriorating bridge
system. The framework assumes that dynamic response data are obtained in a
sequential fashion from deployed accelerometers, subsequently processed by an
output-only operational modal analysis scheme for identifying the system's
modal characteristics. We employ a classical Bayesian model updating
methodology to sequentially learn the deterioration and estimate the structural
damage evolution over time. This leads to sequential updating of the structural
reliability, which constitutes the basis for a preposterior Bayesian decision
analysis. Alternative actions are defined and a heuristic-based approach is
employed for the life-cycle optimization. By solving the preposterior Bayesian
decision analysis, one is able to quantify the benefit of the availability of
long-term SHM vibrational data. Numerical investigations show that this
framework can provide quantitative measures on the optimality of an SHM system
in a specific decision context.
| stat |
Generalized Pearson correlation squares for capturing mixtures of
bivariate linear dependences | Motivated by the pressing needs for capturing complex but interpretable
variable relationships in scientific research, here we generalize the squared
Pearson correlation to capture a mixture of linear dependences between two
real-valued random variables, with or without an index variable that specifies
the line memberships. We construct generalized Pearson correlation squares by
focusing on three aspects: the exchangeability of the two variables, the
independence of parametric model assumptions, and the availability of
population-level parameters. For the computation of the generalized Pearson
correlation square from a sample without line-membership specification, we
develop a K-lines clustering algorithm, where K, the number of lines, can be
chosen in a data-adaptive way. With our defined population-level generalized
Pearson correlation squares, we derive the asymptotic distributions of the
sample-level statistics to enable efficient statistical inference. Simulation
studies verify the theoretical results and compare the generalized Pearson
correlation squares with other widely-used association measures in terms of
power. Gene expression data analysis demonstrates the effectiveness of the
generalized Pearson correlation squares in capturing interpretable gene-gene
relationships missed by other measures. We implement the estimation and
inference procedures in an R package gR2.
| stat |
Diagonally-Dominant Principal Component Analysis | We consider the problem of decomposing a large covariance matrix into the sum
of a low-rank matrix and a diagonally dominant matrix, and we call this problem
the "Diagonally-Dominant Principal Component Analysis (DD-PCA)". DD-PCA is an
effective tool for designing statistical methods for strongly correlated data.
We showcase the use of DD-PCA in two statistical problems: covariance matrix
estimation, and global detection in multiple testing. Using the output of
DD-PCA, we propose a new estimator for estimating a large covariance matrix
with factor structure. Thanks to a nice property of diagonally dominant
matrices, this estimator enjoys the advantage of simultaneous good estimation
of the covariance matrix and the precision matrix (by a plain inversion). A
plug-in of this estimator to linear discriminant analysis and portfolio
optimization yields appealing performance in real data. We also propose two new
tests for testing the global null hypothesis in multiple testing when the
$z$-scores have a factor covariance structure. Both tests first use DD-PCA to
adjust the individual $p$-values and then plug in the adjusted $p$-values to
the Higher Criticism (HC) test. These new tests significantly improve over the
HC test and compare favorably with other existing tests. For computation of
DD-PCA, we propose an iterative projection algorithm and an ADMM algorithm.
| stat |
Novel Bayesian Procrustes Variance-based Inferences in Geometric
Morphometrics & Novel R package: BPviGM1 | Compared to abundant classical statistics-based literature, to date, very
little Bayesian literature exists on Procrustes shape analysis in Geometric
Morphometrics, probably because of being a relatively new branch of statistical
research and because of inherent computational difficulty associated with
Bayesian analysis. Moreover, we may obtain a plethora of novel inferences from
Bayesian Procrustes analysis of shape parameter distributions. In this paper,
we propose to regard the posterior of Procrustes shape variance as
morphological variability indicators. Here we propose novel Bayesian
methodologies for Procrustes shape analysis based on landmark data's isotropic
variance assumption and propose a Bayesian statistical test for model
validation of new species discovery using morphological variation reflected in
the posterior distribution of landmark-variance of objects studied under
Geometric Morphometrics. We will consider Gaussian distribution-based and
heavy-tailed t distribution-based models for Procrustes analysis.
To date, we are not aware of any direct R package for Bayesian Procrustes
analysis for landmark-based Geometric Morphometrics. Hence, we introduce a
novel, simple R package \textbf{BPviGM1} ("Bayesian Procrustes Variance-based
inferences in Geometric Morphometrics 1"), which essentially contains the R
code implementations of the computations for proposed models and methodologies,
such as R function for Markov Chain Monte Carlo (MCMC) run for drawing samples
from posterior of parameters of concern and R function for the proposed
Bayesian test of model validation based on significance morphological
variation. As an application, we can quantitatively show that primate male-face
may be genetically viable to more shape-variation than the same for females.
| stat |
Sequential importance sampling for multi-resolution Kingman-Tajima
coalescent counting | Statistical inference of evolutionary parameters from molecular sequence data
relies on coalescent models to account for the shared genealogical ancestry of
the samples. However, inferential algorithms do not scale to available data
sets. A strategy to improve computational efficiency is to rely on simpler
coalescent and mutation models, resulting in smaller hidden state spaces. An
estimate of the cardinality of the state-space of genealogical trees at
different resolutions is essential to decide the best modeling strategy for a
given dataset. To our knowledge, there is neither an exact nor approximate
method to determine these cardinalities. We propose a sequential importance
sampling algorithm to estimate the cardinality of the space of genealogical
trees under different coalescent resolutions. Our sampling scheme proceeds
sequentially across the set of combinatorial constraints imposed by the data.
We analyse the cardinality of different genealogical tree spaces on simulations
to study the settings that favor coarser resolutions. We estimate the
cardinality of genealogical tree spaces from mtDNA data from the 1000 genomes
and a sample from a Melanesian population to illustrate the settings in which
it is advantageous to employ coarser resolutions.
| stat |
Fractional ridge regression: a fast, interpretable reparameterization of
ridge regression | Ridge regression (RR) is a regularization technique that penalizes the
L2-norm of the coefficients in linear regression. One of the challenges of
using RR is the need to set a hyperparameter ($\alpha$) that controls the
amount of regularization. Cross-validation is typically used to select the best
$\alpha$ from a set of candidates. However, efficient and appropriate selection
of $\alpha$ can be challenging, particularly where large amounts of data are
analyzed. Because the selected $\alpha$ depends on the scale of the data and
predictors, it is not straightforwardly interpretable. Here, we propose to
reparameterize RR in terms of the ratio $\gamma$ between the L2-norms of the
regularized and unregularized coefficients. This approach, called fractional RR
(FRR), has several benefits: the solutions obtained for different $\gamma$ are
guaranteed to vary, guarding against wasted calculations, and automatically
span the relevant range of regularization, avoiding the need for arduous manual
exploration. We provide an algorithm to solve FRR, as well as open-source
software implementations in Python and MATLAB
(https://github.com/nrdg/fracridge). We show that the proposed method is fast
and scalable for large-scale data problems, and delivers results that are
straightforward to interpret and compare across models and datasets.
| stat |
A Conway-Maxwell-Multinomial Distribution for Flexible Modeling of
Clustered Categorical Data | Categorical data are often observed as counts resulting from a fixed number
of trials in which each trial consists of making one selection from a
prespecified set of categories. The multinomial distribution serves as a
standard model for such clustered data but assumes that trials are independent
and identically distributed. Extensions such as Dirichlet-multinomial and
random-clumped multinomial can express positive association, where trials are
more likely to result in a common category due to membership in a common
cluster. This work considers a Conway-Maxwell-multinomial (CMM) distribution
for modeling clustered categorical data exhibiting positively or negatively
associated trials. The CMM distribution features a dispersion parameter which
allows it to adapt to a range of association levels and includes several
recognizable distributions as special cases. We explore properties of CMM,
illustrate its flexible characteristics, identify a method to efficiently
compute maximum likelihood (ML) estimates, present simulations of small sample
properties under ML estimation, and demonstrate the model via several data
analysis examples.
| stat |
Generalization Error Bounds with Probabilistic Guarantee for SGD in
Nonconvex Optimization | The success of deep learning has led to a rising interest in the
generalization property of the stochastic gradient descent (SGD) method, and
stability is one popular approach to study it. Existing works based on
stability have studied nonconvex loss functions, but only considered the
generalization error of the SGD in expectation. In this paper, we establish
various generalization error bounds with probabilistic guarantee for the SGD.
Specifically, for both general nonconvex loss functions and gradient dominant
loss functions, we characterize the on-average stability of the iterates
generated by SGD in terms of the on-average variance of the stochastic
gradients. Such characterization leads to improved bounds for the
generalization error for SGD. We then study the regularized risk minimization
problem with strongly convex regularizers, and obtain improved generalization
error bounds for proximal SGD. With strongly convex regularizers, we further
establish the generalization error bounds for nonconvex loss functions under
proximal SGD with high-probability guarantee, i.e., exponential concentration
in probability.
| stat |
Separable nonlinear least-squares parameter estimation for complex
dynamic systems | Nonlinear dynamic models are widely used for characterizing functional forms
of processes that govern complex biological pathway systems. Over the past
decade, validation and further development of these models became possible due
to data collected via high-throughput experiments using methods from molecular
biology. While these data are very beneficial, they are typically incomplete
and noisy, so that inferring parameter values for complex dynamic models is
associated with serious computational challenges. Fortunately, many biological
systems have embedded linear mathematical features, which may be exploited,
thereby improving fits and leading to better convergence of optimization
algorithms.
In this paper, we explore options of inference for dynamic models using a
novel method of {\it separable nonlinear least-squares optimization}, and
compare its performance to the traditional nonlinear least-squares method. The
numerical results from extensive simulations suggest that the proposed approach
is at least as accurate as the traditional nonlinear least-squares, but usually
superior, while also enjoying a substantial reduction in computational time.
| stat |
A Novel Method of Marginalisation using Low Discrepancy Sequences for
Integrated Nested Laplace Approximations | Recently, it has been shown that approximations to marginal posterior
distributions obtained using a low discrepancy sequence (LDS) can outperform
standard grid-based methods with respect to both accuracy and computational
efficiency. This recent method, which we will refer to as LDS-StM, can also
produce good approximations to multimodal posteriors. However, implementation
of LDS-StM into integrated nested Laplace approximations (INLA), a methodology
in which grid-based methods are used, is challenging. Motivated by this
problem, we propose modifications to LDS-StM that improves the approximations
and make it compatible with INLA, without sacrificing computational speed. We
also present two examples to demonstrate that LDS-StM with modifications can
outperform INLA's own grid approximation with respect to speed and accuracy. We
also demonstrate the flexibility of the new approach for the approximation of
multimodal marginals.
| stat |
Semi-parametric Bayesian variable selection for gene-environment
interactions | Many complex diseases are known to be affected by the interactions between
genetic variants and environmental exposures beyond the main genetic and
environmental effects. Study of gene-environment (G$\times$E) interactions is
important for elucidating the disease etiology. Existing Bayesian methods for
G$\times$E interaction studies are challenged by the high-dimensional nature of
the study and the complexity of environmental influences. Many studies have
shown the advantages of penalization methods in detecting G$\times$E
interactions in "large p, small n" settings. However, Bayesian variable
selection, which can provide fresh insight into G$\times$E study, has not been
widely examined. We propose a novel and powerful semi-parametric Bayesian
variable selection model that can investigate linear and nonlinear G$\times$E
interactions simultaneously. Furthermore, the proposed method can conduct
structural identification by distinguishing nonlinear interactions from
main-effects-only case within the Bayesian framework. Spike and slab priors are
incorporated on both individual and group levels to identify the sparse main
and interaction effects. The proposed method conducts Bayesian variable
selection more efficiently than existing methods. Simulation shows that the
proposed model outperforms competing alternatives in terms of both
identification and prediction. The proposed Bayesian method leads to the
identification of main and interaction effects with important implications in a
high-throughput profiling study with high-dimensional SNP data.
| stat |
Variational Learning of Individual Survival Distributions | The abundance of modern health data provides many opportunities for the use
of machine learning techniques to build better statistical models to improve
clinical decision making. Predicting time-to-event distributions, also known as
survival analysis, plays a key role in many clinical applications. We introduce
a variational time-to-event prediction model, named Variational Survival
Inference (VSI), which builds upon recent advances in distribution learning
techniques and deep neural networks. VSI addresses the challenges of
non-parametric distribution estimation by ($i$) relaxing the restrictive
modeling assumptions made in classical models, and ($ii$) efficiently handling
the censored observations, {\it i.e.}, events that occur outside the
observation window, all within the variational framework. To validate the
effectiveness of our approach, an extensive set of experiments on both
synthetic and real-world datasets is carried out, showing improved performance
relative to competing solutions.
| stat |
Mean-Squared Accuracy of Good-Turing Estimator | The brilliant method due to Good and Turing allows for estimating objects not
occurring in a sample. The problem, known under names "sample coverage" or
"missing mass" goes back to their cryptographic work during WWII, but over
years has found has many applications, including language modeling, inference
in ecology and estimation of distribution properties. This work characterizes
the maximal mean-squared error of the Good-Turing estimator, for any sample
\emph{and} alphabet size.
| stat |
Identification and Estimation of Heterogeneous Treatment Effects under
Non-compliance or Non-ignorable assignment | We provide sufficient conditions for the identification of the heterogeneous
treatment effects, defined as the conditional expectation for the differences
of potential outcomes given the untreated outcome, under the nonignorable
treatment condition and availability of the information on the marginal
distribution of the untreated outcome. These functions are useful both to
identify the average treatment effects (ATE) and to determine the treatment
assignment policy. The identification holds in the following two general setups
prevalent in applied studies: (i) a randomized controlled trial with one-sided
noncompliance and (ii) an observational study with nonignorable assignment with
the information on the marginal distribution of the untreated outcome or its
sample moments. To handle the setup with many integrals and missing values, we
propose a (quasi-)Bayesian estimation method for HTE and ATE and examine its
properties through simulation studies. We also apply the proposed method to the
dataset obtained by the National Job Training Partnership Act Study.
| stat |
Lift Up and Act! Classifier Performance in Resource-Constrained
Applications | Classification tasks are common across many fields and applications where the
decision maker's action is limited by resource constraints. In direct marketing
only a subset of customers is contacted; scarce human resources limit the
number of interviews to the most promising job candidates; limited donated
organs are prioritized to those with best fit. In such scenarios, performance
measures such as the classification matrix, ROC analysis, and even ranking
metrics such as AUC measures outcomes different from the action of interest. At
the same time, gains and lift that do measure the relevant outcome are rarely
used by machine learners. In this paper we define resource-constrained
classifier performance as a task distinguished from classification and ranking.
We explain how gains and lift can lead to different algorithm choices and
discuss the effect of class distribution.
| stat |
Exact Bayesian inference for diffusion driven Cox processes | In this paper we present a novel methodology to perform Bayesian inference
for Cox processes in which the intensity function is driven by a diffusion
process. The novelty lies on the fact that no discretisation error is involved,
despite the non-tractability of both the likelihood function and the transition
density of the diffusion. The methodology is based on an MCMC algorithm and its
exactness is built on retrospective sampling techniques. The efficiency of the
methodology is investigated in some simulated examples and its applicability is
illustrated in some real data analyses.
| stat |
Pathogen.jl: Infectious Disease Transmission Network Modelling with
Julia | We introduce Pathogen.jl for simulation and inference of transmission network
individual level models (TN-ILMs) of infectious disease spread. TN-ILMs can be
used to jointly infer transmission networks, event times, and model parameters
within a Bayesian framework via Markov chain Monte Carlow (MCMC). We detail our
specific strategies for conducting MCMC for TN-ILMs, and our implementation of
these strategies in the Julia package, Pathogen.jl, which leverages key
features of the Julia language. We provide an example using Pathogen.jl to
simulate an epidemic following a susceptible-infectious-removed (SIR) TN-ILM,
and then perform inference using observations that were generated from that
epidemic. We also demonstrate the functionality of Pathogen.jl with an
application of TN-ILMs to data from a measles outbreak that occurred in
Hagelloch, Germany in 1861 (Pfeilsticker 1863; Oesterle 1992).
| stat |
Spectral Density Estimation for Random Fields via Periodic Embeddings | We introduce methods for estimating the spectral density of a random field on
a $d$-dimensional lattice from incomplete gridded data. Data are iteratively
imputed onto an expanded lattice according to a model with a periodic
covariance function. The imputations are convenient computationally, in that
circulant embedding and preconditioned conjugate gradient methods can produce
imputations in $O(n\log n)$ time and $O(n)$ memory. However, these so-called
periodic imputations are motivated mainly by their ability to produce accurace
spectral density estimates. In addition, we introduce a parametric filtering
method that is designed to reduce periodogram smoothing bias. The paper
contains theoretical results studying properties of the imputed data
periodogram and numerical and simulation studies comparing the performance of
the proposed methods to existing approaches in a number of scenarios. We
present an application to a gridded satellite surface temperature dataset with
missing values.
| stat |
New frontiers in Bayesian modeling using the INLA package in R | The INLA package provides a tool for computationally efficient Bayesian
modeling and inference for various widely used models, more formally the class
of latent Gaussian models. It is a non-sampling based framework which provides
approximate results for Bayesian inference, using sparse matrices. The swift
uptake of this framework for Bayesian modeling is rooted in the computational
efficiency of the approach and catalyzed by the demand presented by the big
data era. In this paper, we present new developments within the INLA package
with the aim to provide a computationally efficient mechanism for the Bayesian
inference of relevant challenging situations.
| stat |
Exact Monte Carlo likelihood-based inference for jump-diffusion
processes | Statistical inference for discretely observed jump-diffusion processes is a
complex problem which motivates new methodological challenges. Thus existing
approaches invariably resort to time-discretisations which inevitably lead to
approximations in inference. In this paper, we give the first general
collection of methodologies for exact (in this context meaning
discretisation-free) likelihood-based inference for discretely observed
jump-diffusions. We shall introduce both frequentist and Bayesian approaches,
illustrating the methodology through simulated and real examples.
| stat |
rFIA: An R package for estimation of forest attributes with the Forest
Inventory and Analysis Database | Forest Inventory and Analysis (FIA) is a US Department of Agriculture Forest
Service program that aims to monitor changes in forests across the US. FIA
hosts one of the largest ecological datasets in the world, though its
complexity limits access for many potential users. rFIA is an R package
designed to simplify the estimation of forest attributes using data collected
by the FIA Program. Specifically, rFIA improves access to the spatio-temporal
estimation capacity of the FIA Database via space-time indexed summaries of
forest variables within user-defined population boundaries (e.g., geographic,
temporal, biophysical). The package implements multiple design-based
estimators, and has been validated against official estimates and sampling
errors produced by the FIA Program. We demonstrate the utility of rFIA by
assessing changes in abundance and mortality rates of ash (Fraxinus spp.)
populations in the Lower Peninsula of Michigan following the establishment of
emerald ash borer (Agrilus planipennis).
| stat |
A representer theorem for deep neural networks | We propose to optimize the activation functions of a deep neural network by
adding a corresponding functional regularization to the cost function. We
justify the use of a second-order total-variation criterion. This allows us to
derive a general representer theorem for deep neural networks that makes a
direct connection with splines and sparsity. Specifically, we show that the
optimal network configuration can be achieved with activation functions that
are nonuniform linear splines with adaptive knots. The bottom line is that the
action of each neuron is encoded by a spline whose parameters (including the
number of knots) are optimized during the training procedure. The scheme
results in a computational structure that is compatible with the existing
deep-ReLU, parametric ReLU, APL (adaptive piecewise-linear) and MaxOut
architectures. It also suggests novel optimization challenges, while making the
link with $\ell_1$ minimization and sparsity-promoting techniques explicit.
| stat |
Proper scoring rules for evaluating asymmetry in density forecasting | This paper proposes a novel asymmetric continuous probabilistic score (ACPS)
for evaluating and comparing density forecasts. It extends the proposed score
and defines a weighted version, which emphasizes regions of interest, such as
the tails or the center of a variable's range. A test is also introduced to
statistically compare the predictive ability of different forecasts. The ACPS
is of general use in any situation where the decision maker has asymmetric
preferences in the evaluation of the forecasts. In an artificial experiment,
the implications of varying the level of asymmetry in the ACPS are illustrated.
Then, the proposed score and test are applied to assess and compare density
forecasts of macroeconomic relevant datasets (US employment growth) and of
commodity prices (oil and electricity prices) with particular focus on the
recent COVID-19 crisis period.
| stat |
Analysis of regression discontinuity designs using censored data | In medical settings, treatment assignment may be determined by a clinically
important covariate that predicts patients' risk of event. There is a class of
methods from the social science literature known as regression discontinuity
(RD) designs that can be used to estimate the treatment effect in this
situation. Under certain assumptions, such an estimand enjoys a causal
interpretation. However, few authors have discussed the use of RD for censored
data. In this paper, we show how to estimate causal effects under the
regression discontinuity design for censored data. The proposed estimation
procedure employs a class of censoring unbiased transformations that includes
inverse probability censored weighting and doubly robust transformation
schemes. Simulation studies demonstrate the utility of the proposed
methodology.
| stat |
Aligning Subjective Ratings in Clinical Decision Making | In addition to objective indicators (e.g. laboratory values), clinical data
often contain subjective evaluations by experts (e.g. disease severity
assessments). While objective indicators are more transparent and robust, the
subjective evaluation contains a wealth of expert knowledge and intuition. In
this work, we demonstrate the potential of pairwise ranking methods to align
the subjective evaluation with objective indicators, creating a new score that
combines their advantages and facilitates diagnosis. In a case study on
patients at risk for developing Psoriatic Arthritis, we illustrate that the
resulting score (1) increases classification accuracy when detecting disease
presence/absence, (2) is sparse and (3) provides a nuanced assessment of
severity for subsequent analysis.
| stat |
Adjustment for Biased Sampling Using NHANES Derived Propensity Weights | The Consent-to-Contact (C2C) registry at the University of California, Irvine
collects data from community participants to aid in the recruitment to clinical
research studies. Self-selection into the C2C likely leads to bias due in part
to enrollees having more years of education relative to the US general
population. Salazar et al. (2020) recently used the C2C to examine associations
of race/ethnicity with participant willingness to be contacted about research
studies. To address questions about generalizability of estimated associations
we estimate propensity for self-selection into the convenience sample weights
using data from the National Health and Nutrition Examination Survey (NHANES).
We create a combined dataset of C2C and NHANES subjects and compare different
approaches (logistic regression, covariate balancing propensity score, entropy
balancing, and random forest) for estimating the probability of membership in
C2C relative to NHANES. We propose methods to estimate the variance of
parameter estimates that account for uncertainty that arises from estimating
propensity weights. Simulation studies explore the impact of propensity weight
estimation on uncertainty. We demonstrate the approach by repeating the
analysis by Salazar et al. with the deduced propensity weights for the C2C
subjects and contrast the results of the two analyses. This method can be
implemented using our estweight package in R available on GitHub.
| stat |
LqRT: Robust Hypothesis Testing of Location Parameters using
Lq-Likelihood-Ratio-Type Test in Python | A t-test is considered a standard procedure for inference on population means
and is widely used in scientific discovery. However, as a special case of a
likelihood-ratio test, t-test often shows drastic performance degradation due
to the deviations from its hard-to-verify distributional assumptions.
Alternatively, in this article, we propose a new two-sample
Lq-likelihood-ratio-type test (LqRT) along with an easy-to-use Python package
for implementation. LqRT preserves high power when the distributional
assumption is violated, and maintains the satisfactory performance when the
assumption is valid. As numerical studies suggest, LqRT dominates many other
robust tests in power, such as Wilcoxon test and sign test, while maintaining a
valid size. To the extent that the robustness of the Wilcoxon test (minimum
asymptotic relative efficiency (ARE) of the Wilcoxon test vs the t-test is
0.864) suggests that the Wilcoxon test should be the default test of choice
(rather than "use Wilcoxon if there is evidence of non-normality", the default
position should be "use Wilcoxon unless there is good reason to believe the
normality assumption"), the results in this article suggest that the LqRT is
potentially the new default go-to test for practitioners.
| stat |
Orbit: Probabilistic Forecast with Exponential Smoothing | Time series forecasting is an active research topic in academia as well as
industry. Although we see an increasing amount of adoptions of machine learning
methods in solving some of those forecasting challenges, statistical methods
remain powerful while dealing with low granularity data. This paper introduces
a refined Bayesian exponential smoothing model with the help of probabilistic
programming languages including Stan. Our model refinements include additional
global trend, transformation for multiplicative form, noise distribution and
choice of priors. A benchmark study is conducted on a rich set of time-series
data sets for our models along with other well-known time series models.
| stat |
Automatic adaptation of MCMC algorithms | Markov chain Monte Carlo (MCMC) methods are ubiquitous tools for
simulation-based inference in many fields but designing and identifying good
MCMC samplers is still an open question. This paper introduces a novel MCMC
algorithm, namely, Auto Adapt MCMC. For sampling variables or blocks of
variables, we use two levels of adaptation where the inner adaptation optimizes
the MCMC performance within each sampler, while the outer adaptation explores
the valid space of kernels to find the optimal samplers. We provide a
theoretical foundation for our approach. To show the generality and usefulness
of the approach, we describe a framework using only standard MCMC samplers as
candidate samplers and some adaptation schemes for both inner and outer
iterations. In several benchmark problems, we show that our proposed approach
substantially outperforms other approaches, including an automatic blocking
algorithm, in terms of MCMC efficiency and computational time.
| stat |
Integrable Nonparametric Flows | We introduce a method for reconstructing an infinitesimal normalizing flow
given only an infinitesimal change to a (possibly unnormalized) probability
distribution. This reverses the conventional task of normalizing flows --
rather than being given samples from a unknown target distribution and learning
a flow that approximates the distribution, we are given a perturbation to an
initial distribution and aim to reconstruct a flow that would generate samples
from the known perturbed distribution. While this is an underdetermined
problem, we find that choosing the flow to be an integrable vector field yields
a solution closely related to electrostatics, and a solution can be computed by
the method of Green's functions. Unlike conventional normalizing flows, this
flow can be represented in an entirely nonparametric manner. We validate this
derivation on low-dimensional problems, and discuss potential applications to
problems in quantum Monte Carlo and machine learning.
| stat |
Slalom: Fast, Verifiable and Private Execution of Neural Networks in
Trusted Hardware | As Machine Learning (ML) gets applied to security-critical or sensitive
domains, there is a growing need for integrity and privacy for outsourced ML
computations. A pragmatic solution comes from Trusted Execution Environments
(TEEs), which use hardware and software protections to isolate sensitive
computations from the untrusted software stack. However, these isolation
guarantees come at a price in performance, compared to untrusted alternatives.
This paper initiates the study of high performance execution of Deep Neural
Networks (DNNs) in TEEs by efficiently partitioning DNN computations between
trusted and untrusted devices. Building upon an efficient outsourcing scheme
for matrix multiplication, we propose Slalom, a framework that securely
delegates execution of all linear layers in a DNN from a TEE (e.g., Intel SGX
or Sanctum) to a faster, yet untrusted, co-located processor. We evaluate
Slalom by running DNNs in an Intel SGX enclave, which selectively delegates
work to an untrusted GPU. For canonical DNNs (VGG16, MobileNet and ResNet
variants) we obtain 6x to 20x increases in throughput for verifiable inference,
and 4x to 11x for verifiable and private inference.
| stat |
Anomaly Detection in High Dimensional Data | The HDoutliers algorithm is a powerful unsupervised algorithm for detecting
anomalies in high-dimensional data, with a strong theoretical foundation.
However, it suffers from some limitations that significantly hinder its
performance level, under certain circumstances. In this article, we propose an
algorithm that addresses these limitations. We define an anomaly as an
observation that deviates markedly from the majority with a large distance gap.
An approach based on extreme value theory is used for the anomalous threshold
calculation. Using various synthetic and real datasets, we demonstrate the wide
applicability and usefulness of our algorithm, which we call the stray
algorithm. We also demonstrate how this algorithm can assist in detecting
anomalies present in other data structures using feature engineering. We show
the situations where the stray algorithm outperforms the HDoutliers algorithm
both in accuracy and computational time. This framework is implemented in the
open source R package stray.
| stat |
What is the best predictor that you can compute in five minutes using a
given Bayesian hierarchical model? | The goal of this paper is to provide a way for statisticians to answer the
question posed in the title of this article using any Bayesian hierarchical
model of their choosing and without imposing additional restrictive model
assumptions. We are motivated by the fact that the rise of ``big data'' has
created difficulties for statisticians to directly apply their methods to big
datasets. We introduce a ``data subset model'' to the popular ``data model,
process model, and parameter model'' framework used to summarize Bayesian
hierarchical models. The hyperparameters of the data subset model are specified
constructively in that they are chosen such that the implied size of the subset
satisfies pre-defined computational constraints. Thus, these hyperparameters
effectively calibrates the statistical model to the computer itself to obtain
predictions/estimations in a pre-specified amount of time. Several properties
of the data subset model are provided including: propriety, partial
sufficiency, and semi-parametric properties. Furthermore, we show that subsets
of normally distributed data are asymptotically partially sufficient under
reasonable constraints. Results from a simulated dataset will be presented
across different computers, to show the effect of the computer on the
statistical analysis. Additionally, we provide a joint spatial analysis of two
different environmental datasets.
| stat |
Asymptotically Optimal Strategies For Combinatorial Semi-Bandits in
Polynomial Time | We consider combinatorial semi-bandits with uncorrelated Gaussian rewards. In
this article, we propose the first method, to the best of our knowledge, that
enables to compute the solution of the Graves-Lai optimization problem in
polynomial time for many combinatorial structures of interest. In turn, this
immediately yields the first known approach to implement asymptotically optimal
algorithms in polynomial time for combinatorial semi-bandits.
| stat |
An R Autograder for PrairieLearn | We describe how we both use and extend the PrarieLearn framework by taking
advantage of its built-in support for external auto-graders. By using a custom
Docker container, we can match our course requirements perfectly. Moreover, by
relying on the flexibility of the interface we can customize our Docker
container. A specific extension for unit testing is described which creates
context-dependent difference between student answers and reference solution
providing a more comprehensive response at test time.
| stat |
Score Engineered Robust Least Squares Regression | In other FICO Technical Papers, I have shown how to fit Generalized Additive
Models (GAM) with shape constraints using quadratic programming applied to
B-Spline component functions. In this paper, I extend the method to Robust
Least Squares Regression.
| stat |
Adaptive Partitioning Design and Analysis for Emulation of a Complex
Computer Code | Computer models are used as replacements for physical experiments in a large
variety of applications. Nevertheless, direct use of the computer model for the
ultimate scientific objective is often limited by the complexity and cost of
the model. Historically, Gaussian process regression has proven to be the
almost ubiquitous choice for a fast statistical emulator for such a computer
model, due to its flexible form and analytical expressions for measures of
predictive uncertainty. However, even this statistical emulator can be
computationally intractable for large designs, due to computing time increasing
with the cube of the design size. Multiple methods have been proposed for
addressing this problem. We discuss several of them, and compare their
predictive and computational performance in several scenarios.
We then propose solving this problem using an adaptive partitioning emulator
(APE). The new approach is motivated by the idea that most computer models are
only complex in particular regions of the input space. By taking a
data-adaptive approach to the development of a design, and choosing to
partition the space in the regions of highest variability, we obtain a higher
density of points in these regions and hence accurate prediction.
| stat |
Clustering Higher Order Data: An Application to Pediatric Multi-variable
Longitudinal Data | Physical activity levels are an important predictor of cardiovascular health
and increasingly being measured by sensors, like accelerometers. Accelerometers
produce rich multivariate data that can inform important clinical decisions
related to individual patients and public health. The CHAMPION study, a study
of youth with chronic inflammatory conditions, aims to determine the links
between heart health, inflammation, physical activity, and fitness. The
accelerometer data from CHAMPION is represented as 4-dimensional arrays, and a
finite mixture of multidimensional arrays model is developed for clustering.
The use of model-based clustering for multidimensional arrays has thus far been
limited to two-dimensional arrays, i.e., matrices or order-two tensors, and the
work in this paper can also be seen as an approach for clustering D-dimensional
arrays for D > 2 or, in other words, for clustering order-D tensors.
| stat |
Multi-View Matrix Completion for Multi-Label Image Classification | There is growing interest in multi-label image classification due to its
critical role in web-based image analytics-based applications, such as
large-scale image retrieval and browsing. Matrix completion has recently been
introduced as a method for transductive (semi-supervised) multi-label
classification, and has several distinct advantages, including robustness to
missing data and background noise in both feature and label space. However, it
is limited by only considering data represented by a single-view feature, which
cannot precisely characterize images containing several semantic concepts. To
utilize multiple features taken from different views, we have to concatenate
the different features as a long vector. But this concatenation is prone to
over-fitting and often leads to very high time complexity in MC based image
classification. Therefore, we propose to weightedly combine the MC outputs of
different views, and present the multi-view matrix completion (MVMC) framework
for transductive multi-label image classification. To learn the view
combination weights effectively, we apply a cross validation strategy on the
labeled set. In the learning process, we adopt the average precision (AP) loss,
which is particular suitable for multi-label image classification. A least
squares loss formulation is also presented for the sake of efficiency, and the
robustness of the algorithm based on the AP loss compared with the other losses
is investigated. Experimental evaluation on two real world datasets (PASCAL
VOC' 07 and MIR Flickr) demonstrate the effectiveness of MVMC for transductive
(semi-supervised) multi-label image classification, and show that MVMC can
exploit complementary properties of different features and output-consistent
labels for improved multi-label image classification.
| stat |
Empirical Bayes Model Averaging with Influential Observations: Tuning
Zellner's g Prior for Predictive Robustness | We investigate the behavior of Bayesian model averaging (BMA) for the normal
linear regression model in the presence of influential observations that
contribute to model misfit, and propose remedies to attenuate the potential
negative impacts of such observations on inference and prediction. The
methodology is motivated by the view that well-behaved residuals and good
predictive performance often go hand-in-hand. The study focuses on regression
models that use variants on Zellner's g prior. By studying the impact of
various forms of model misfit on BMA predictions in simple situations we
identify prescriptive guidelines for "tuning" Zellner's g prior to obtain
optimal predictions. The tuning of the prior distribution is obtained by
considering theoretical properties that should be enjoyed by the optimal fits
of the various models in the BMA ensemble. The methodology can be thought of as
an "empirical Bayes" approach to modeling, as the data help to inform the
specification of the prior in an attempt to attenuate the negative impact of
influential cases.
| stat |
Using the "Hidden" Genome to Improve Classification of Cancer Types | It is increasingly common clinically for cancer specimens to be examined
using techniques that identify somatic mutations. In principle these mutational
profiles can be used to diagnose the tissue of origin, a critical task for the
3-5% of tumors that have an unknown primary site. Diagnosis of primary site is
also critical for screening tests that employ circulating DNA. However, most
mutations observed in any new tumor are very rarely occurring mutations, and
indeed the preponderance of these may never have been observed in any previous
recorded tumor. To create a viable diagnostic tool we need to harness the
information content in this "hidden genome" of variants for which no direct
information is available. To accomplish this we propose a multi-level
meta-feature regression to extract the critical information from rare variants
in the training data in a way that permits us to also extract diagnostic
information from any previously unobserved variants in the new tumor sample. A
scalable implementation of the model is obtained by combining a
high-dimensional feature screening approach with a group-lasso penalized
maximum likelihood approach based on an equivalent mixed-effect representation
of the multilevel model. We apply the method to the Cancer Genome Atlas
whole-exome sequencing data set including 3702 tumor samples across 7 common
cancer sites. Results show that our multi-level approach can harness
substantial diagnostic information from the hidden genome.
| stat |
Kernel embedded nonlinear observational mappings in the variational
mapping particle filter | Recently, some works have suggested methods to combine variational
probabilistic inference with Monte Carlo sampling. One promising approach is
via local optimal transport. In this approach, a gradient steepest descent
method based on local optimal transport principles is formulated to transform
deterministically point samples from an intermediate density to a posterior
density. The local mappings that transform the intermediate densities are
embedded in a reproducing kernel Hilbert space (RKHS). This variational mapping
method requires the evaluation of the log-posterior density gradient and
therefore the adjoint of the observational operator. In this work, we evaluate
nonlinear observational mappings in the variational mapping method using two
approximations that avoid the adjoint, an ensemble based approximation in which
the gradient is approximated by the particle covariances in the state and
observational spaces the so-called ensemble space and an RKHS approximation in
which the observational mapping is embedded in an RKHS and the gradient is
derived there. The approximations are evaluated for highly nonlinear
observational operators and in a low-dimensional chaotic dynamical system. The
RKHS approximation is shown to be highly successful and superior to the
ensemble approximation.
| stat |
A Vector Autoregression Prediction Model for COVID-19 Outbreak | Since two people came down a county of north Seattle with positive COVID-19
(coronavirus-19) in 2019, the current total cases in the United States (U.S.)
are over 12 million. Predicting the pandemic trend under effective variables is
crucial to help find a way to control the epidemic. Based on available
literature, we propose a validated Vector Autoregression (VAR) time series
model to predict the positive COVID-19 cases. A real data prediction for U.S.
is provided based on the U.S. coronavirus data. The key message from our study
is that the situation of the pandemic will getting worse if there is no
effective control.
| stat |
Updating Variational Bayes: Fast sequential posterior inference | Variational Bayesian (VB) methods produce posterior inference in a time frame
considerably smaller than traditional Markov Chain Monte Carlo approaches.
Although the VB posterior is an approximation, it has been shown to produce
good parameter estimates and predicted values when a rich classes of
approximating distributions are considered. In this paper we propose Updating
VB (UVB), a recursive algorithm used to update a sequence of VB posterior
approximations in an online setting, with the computation of each posterior
update requiring only the data observed since the previous update. An extension
to the proposed algorithm, named UVB-IS, allows the user to trade accuracy for
a substantial increase in computational speed through the use of importance
sampling. The two methods and their properties are detailed in two separate
simulation studies. Two empirical illustrations of the proposed UVB methods are
provided, including one where a Dirichlet Process Mixture model with a novel
posterior dependence structure is repeatedly updated in the context of
predicting the future behaviour of vehicles on a stretch of the US Highway 101.
| stat |
Extracting more from boosted decision trees: A high energy physics case
study | Particle identification is one of the core tasks in the data analysis
pipeline at the Large Hadron Collider (LHC). Statistically, this entails the
identification of rare signal events buried in immense backgrounds that mimic
the properties of the former. In machine learning parlance, particle
identification represents a classification problem characterized by overlapping
and imbalanced classes. Boosted decision trees (BDTs) have had tremendous
success in the particle identification domain but more recently have been
overshadowed by deep learning (DNNs) approaches. This work proposes an
algorithm to extract more out of standard boosted decision trees by targeting
their main weakness, susceptibility to overfitting. This novel construction
harnesses the meta-learning techniques of boosting and bagging simultaneously
and performs remarkably well on the ATLAS Higgs (H) to tau-tau data set (ATLAS
et al., 2014) which was the subject of the 2014 Higgs ML Challenge
(Adam-Bourdarios et al., 2015). While the decay of Higgs to a pair of tau
leptons was established in 2018 (CMS collaboration et al., 2017) at the
4.9$\sigma$ significance based on the 2016 data taking period, the 2014 public
data set continues to serve as a benchmark data set to test the performance of
supervised classification schemes. We show that the score achieved by the
proposed algorithm is very close to the published winning score which leverages
an ensemble of deep neural networks (DNNs). Although this paper focuses on a
single application, it is expected that this simple and robust technique will
find wider applications in high energy physics.
| stat |
Deep Learning for Energy Markets | Deep Learning is applied to energy markets to predict extreme loads observed
in energy grids. Forecasting energy loads and prices is challenging due to
sharp peaks and troughs that arise due to supply and demand fluctuations from
intraday system constraints. We propose deep spatio-temporal models and extreme
value theory (EVT) to capture theses effects and in particular the tail
behavior of load spikes. Deep LSTM architectures with ReLU and $\tanh$
activation functions can model trends and temporal dependencies while EVT
captures highly volatile load spikes above a pre-specified threshold. To
illustrate our methodology, we use hourly price and demand data from 4719 nodes
of the PJM interconnection, and we construct a deep predictor. We show that
DL-EVT outperforms traditional Fourier time series methods, both in-and
out-of-sample, by capturing the observed nonlinearities in prices. Finally, we
conclude with directions for future research.
| stat |
A Formalization of The Natural Gradient Method for General Similarity
Measures | In optimization, the natural gradient method is well-known for likelihood
maximization. The method uses the Kullback-Leibler divergence, corresponding
infinitesimally to the Fisher-Rao metric, which is pulled back to the parameter
space of a family of probability distributions. This way, gradients with
respect to the parameters respect the Fisher-Rao geometry of the space of
distributions, which might differ vastly from the standard Euclidean geometry
of the parameter space, often leading to faster convergence. However, when
minimizing an arbitrary similarity measure between distributions, it is
generally unclear which metric to use. We provide a general framework that,
given a similarity measure, derives a metric for the natural gradient. We then
discuss connections between the natural gradient method and multiple other
optimization techniques in the literature. Finally, we provide computations of
the formal natural gradient to show overlap with well-known cases and to
compute natural gradients in novel frameworks.
| stat |
A New Approach to Adaptive Data Analysis and Learning via Maximal
Leakage | There is an increasing concern that most current published research findings
are false. The main cause seems to lie in the fundamental disconnection between
theory and practice in data analysis. While the former typically relies on
statistical independence, the latter is an inherently adaptive process: new
hypotheses are formulated based on the outcomes of previous analyses. A recent
line of work tries to mitigate these issues by enforcing constraints, such as
differential privacy, that compose adaptively while degrading gracefully and
thus provide statistical guarantees even in adaptive contexts. Our contribution
consists in the introduction of a new approach, based on the concept of Maximal
Leakage, an information-theoretic measure of leakage of information. The main
result allows us to compare the probability of an event happening when
adaptivity is considered with respect to the non-adaptive scenario. The bound
we derive represents a generalization of the bounds used in non-adaptive
scenarios (e.g., McDiarmid's inequality for $c$-sensitive functions, false
discovery error control via significance level, etc.), and allows us to
replicate or even improve, in certain regimes, the results obtained using
Max-Information or Differential Privacy. In contrast with the line of work
started by Dwork et al., our results do not rely on Differential Privacy but
are, in principle, applicable to every algorithm that has a bounded leakage,
including the differentially private algorithms and the ones with a short
description length.
| stat |
Distributed Generalized Cross-Validation for Divide-and-Conquer Kernel
Ridge Regression and its Asymptotic Optimality | Tuning parameter selection is of critical importance for kernel ridge
regression. To this date, data driven tuning method for divide-and-conquer
kernel ridge regression (d-KRR) has been lacking in the literature, which
limits the applicability of d-KRR for large data sets. In this paper, by
modifying the Generalized Cross-validation (GCV, Wahba, 1990) score, we propose
a distributed Generalized Cross-Validation (dGCV) as a data-driven tool for
selecting the tuning parameters in d-KRR. Not only the proposed dGCV is
computationally scalable for massive data sets, it is also shown, under mild
conditions, to be asymptotically optimal in the sense that minimizing the dGCV
score is equivalent to minimizing the true global conditional empirical loss of
the averaged function estimator, extending the existing optimality results of
GCV to the divide-and-conquer framework.
| stat |
Longitudinal Deep Kernel Gaussian Process Regression | Gaussian processes offer an attractive framework for predictive modeling from
longitudinal data, i.e., irregularly sampled, sparse observations from a set of
individuals over time. However, such methods have two key shortcomings: (i)
They rely on ad hoc heuristics or expensive trial and error to choose the
effective kernels, and (ii) They fail to handle multilevel correlation
structure in the data. We introduce Longitudinal deep kernel Gaussian process
regression (L-DKGPR), which to the best of our knowledge, is the only method to
overcome these limitations by fully automating the discovery of complex
multilevel correlation structure from longitudinal data. Specifically, L-DKGPR
eliminates the need for ad hoc heuristics or trial and error using a novel
adaptation of deep kernel learning that combines the expressive power of deep
neural networks with the flexibility of non-parametric kernel methods. L-DKGPR
effectively learns the multilevel correlation with a novel addictive kernel
that simultaneously accommodates both time-varying and the time-invariant
effects. We derive an efficient algorithm to train L-DKGPR using latent space
inducing points and variational inference. Results of extensive experiments on
several benchmark data sets demonstrate that L-DKGPR significantly outperforms
the state-of-the-art longitudinal data analysis (LDA) methods.
| stat |
A priori estimates for classification problems using neural networks | We consider binary and multi-class classification problems using hypothesis
classes of neural networks. For a given hypothesis class, we use Rademacher
complexity estimates and direct approximation theorems to obtain a priori error
estimates for regularized loss functionals.
| stat |
A Homotopic Method to Solve the Lasso Problems with an Improved Upper
Bound of Convergence Rate | In optimization, it is known that when the objective functions are strictly
convex and well-conditioned, gradient based approaches can be extremely
effective, e.g., achieving the exponential rate in convergence. On the other
hand, the existing Lasso-type of estimator in general cannot achieve the
optimal rate due to the undesirable behavior of the absolute function at the
origin. A homotopic method is to use a sequence of surrogate functions to
approximate the $\ell_1$ penalty that is used in the Lasso-type of estimators.
The surrogate functions will converge to the $\ell_1$ penalty in the Lasso
estimator. At the same time, each surrogate function is strictly convex, which
enables provable faster numerical rate of convergence. In this paper, we
demonstrate that by meticulously defining the surrogate functions, one can
prove faster numerical convergence rate than any existing methods in computing
for the Lasso-type of estimators. Namely, the state-of-the-art algorithms can
only guarantee $O(1/\epsilon)$ or $O(1/\sqrt{\epsilon})$ convergence rates,
while we can prove an $O([\log(1/\epsilon)]^2)$ for the newly proposed
algorithm. Our numerical simulations show that the new algorithm also performs
better empirically.
| stat |
Parallelising MCMC via Random Forests | For Bayesian computation in big data contexts, the divide-and-conquer MCMC
concept splits the whole data set into batches, runs MCMC algorithms separately
over each batch to produce samples of parameters, and combines them to produce
an approximation of the target distribution. In this article, we embed random
forests into this framework and use each subposterior/partial-posterior as a
proposal distribution to implement importance sampling. Unlike the existing
divide-and-conquer MCMC, our methods are based on scaled subposteriors, whose
scale factors are not necessarily restricted to being equal to one or to the
number of subsets. Through several experiments, we show that our methods work
well with models ranging from Gaussian cases to strongly non-Gaussian cases,
and include model misspecification.
| stat |
Deep Distributional Time Series Models and the Probabilistic Forecasting
of Intraday Electricity Prices | Recurrent neural networks (RNNs) with rich feature vectors of past values can
provide accurate point forecasts for series that exhibit complex serial
dependence. We propose two approaches to constructing deep time series
probabilistic models based on a variant of RNN called an echo state network
(ESN). The first is where the output layer of the ESN has stochastic
disturbances and a shrinkage prior for additional regularization. The second
approach employs the implicit copula of an ESN with Gaussian disturbances,
which is a deep copula process on the feature space. Combining this copula with
a non-parametrically estimated marginal distribution produces a deep
distributional time series model. The resulting probabilistic forecasts are
deep functions of the feature vector and also marginally calibrated. In both
approaches, Bayesian Markov chain Monte Carlo methods are used to estimate the
models and compute forecasts. The proposed deep time series models are suitable
for the complex task of forecasting intraday electricity prices. Using data
from the Australian National Electricity Market, we show that our models
provide accurate probabilistic price forecasts. Moreover, the models provide a
flexible framework for incorporating probabilistic forecasts of electricity
demand as additional features. We demonstrate that doing so in the deep
distributional time series model in particular, increases price forecast
accuracy substantially.
| stat |
A likelihood-based approach for cure regression models | We propose a new likelihood-based approach for estimation, inference and
variable selection for parametric cure regression models in time-to-event
analysis under random right-censoring. In this context, it often happens that
some subjects are "cured", i.e., they will never experience the event of
interest. Then, the sample of censored observations is an unlabeled mixture of
cured and "susceptible" subjects. Using inverse probability censoring weighting
(IPCW), we propose a likelihood-based estimation procedure for the cure
regression model without making assumptions about the distribution of survival
times for the susceptible subjects. The IPCW approach does require a
preliminary estimate of the censoring distribution, for which general
parametric, semi- or non-parametric approaches can be used. The incorporation
of a penalty term in our estimation procedure is straightforward; in
particular, we propose L1-type penalties for variable selection. Our
theoretical results are derived under mild assumptions. Simulation experiments
and real data analysis illustrate the effectiveness of the new approach.
| stat |
Deconvolution of dust mixtures by latent Dirichlet allocation in
forensic science | Dust particles recovered from the soles of shoes may be indicative of the
sites recently visited by an individual, and, in particular, of the presence of
an individual at a particular site of interest, e.g., the scene of a crime. By
describing the dust profile of a given site by a multinomial distribution over
a fixed number of dust particle types, we can define the probability
distribution of the mixture of dust recovered from the sole of a shoe via
Latent Dirichlet Allocation. We use Variational Bayesian Inference to study the
parameters of the model, and use their resulting posterior distributions to
make inference on (a) the contributions of sites of interest to a dust mixture,
and (b) the particle profiles associated with these sites.
| stat |
Finding Motifs in Knowledge Graphs using Compression | We introduce a method to find network motifs in knowledge graphs. Network
motifs are useful patterns or meaningful subunits of the graph that recur
frequently. We extend the common definition of a network motif to coincide with
a basic graph pattern. We introduce an approach, inspired by recent work for
simple graphs, to induce these from a given knowledge graph, and show that the
motifs found reflect the basic structure of the graph. Specifically, we show
that in random graphs, no motifs are found, and that when we insert a motif
artificially, it can be detected. Finally, we show the results of motif
induction on three real-world knowledge graphs.
| stat |
Approximate Uncertain Program | Chance constrained program where one seeks to minimize an objective over
decisions which satisfy randomly disturbed constraints with a given probability
is computationally intractable. This paper proposes an approximate approach to
address chance constrained program. Firstly, a single layer neural-network is
used to approximate the function from decision domain to violation probability
domain. The algorithm for updating parameters in single layer neural-network
adopts sequential extreme learning machine. Based on the neural violation
probability approximate model, a randomized algorithm is then proposed to
approach the optimizer in the probabilistic feasible domain of decision. In the
randomized algorithm, samples are extracted from decision domain uniformly at
first. Then, violation probabilities of all samples are calculated according to
neural violation probability approximate model. The ones with violation
probability higher than the required level are discarded. The minimizer in the
remained feasible decision samples is used to update sampling policy. The
policy converges to the optimal feasible decision. Numerical simulations are
implemented to validate the proposed method for non-convex problems comparing
with scenario approach and parallel randomized algorithm. The result shows that
proposed method have improved performance.
| stat |
Sparse regular variation | Regular variation provides a convenient theoretical framework to study large
events. In the multivariate setting, the dependence structure of the positive
extremes is characterized by a measure - the spectral measure - defined on the
positive orthant of the unit sphere. This measure gathers information on the
localization of extreme events and has often a sparse support since severe
events do not simultaneously occur in all directions. However, it is defined
through weak convergence which does not provide a natural way to capture this
sparsity structure.In this paper, we introduce the notion of sparse regular
variation which allows to better learn the dependence structure of extreme
events. This concept is based on the Euclidean projection onto the simplex for
which efficient algorithms are known. We prove that under mild assumptions
sparse regular variation and regular variation are two equivalent notions and
we establish several results for sparsely regularly varying random vectors.
Finally, we illustrate on numerical examples how this new concept allows one to
detect extremal directions.
| stat |
An Eigenmodel for Dynamic Multilayer Networks | Dynamic multilayer networks frequently represent the structure of multiple
co-evolving relations; however, statistical models are not well-developed for
this prevalent network type. Here, we propose a new latent space model for
dynamic multilayer networks. The key feature of our model is its ability to
identify common time-varying structures shared by all layers while also
accounting for layer-wise variation and degree heterogeneity. We establish the
identifiability of the model's parameters and develop a structured mean-field
variational inference approach to estimate the model's posterior, which scales
to networks previously intractable to dynamic latent space models. We
demonstrate the estimation procedure's accuracy and scalability on simulated
networks. We apply the model to two real-world problems: discerning regional
conflicts in a data set of international relations and quantifying infectious
disease spread throughout a school based on the student's daily contact
patterns.
| stat |
Statistical Analysis of Data Repeatability Measures | The advent of modern data collection and processing techniques has seen the
size, scale, and complexity of data grow exponentially. A seminal step in
leveraging these rich datasets for downstream inference is understanding the
characteristics of the data which are repeatable -- the aspects of the data
that are able to be identified under a duplicated analysis. Conflictingly, the
utility of traditional repeatability measures, such as the intraclass
correlation coefficient, under these settings is limited. In recent work, novel
data repeatability measures have been introduced in the context where a set of
subjects are measured twice or more, including: fingerprinting, rank sums, and
generalizations of the intraclass correlation coefficient. However, the
relationships between, and the best practices among these measures remains
largely unknown. In this manuscript, we formalize a novel repeatability
measure, discriminability. We show that it is deterministically linked with the
correlation coefficient under univariate random effect models, and has desired
property of optimal accuracy for inferential tasks using multivariate
measurements. Additionally, we overview and systematically compare
repeatability statistics using both theoretical results and simulations. We
show that the rank sum statistic is deterministically linked to a consistent
estimator of discriminability. The power of permutation tests derived from
these measures are compared numerically under Gaussian and non-Gaussian
settings, with and without simulated batch effects. Motivated by both
theoretical and empirical results, we provide methodological recommendations
for each benchmark setting to serve as a resource for future analyses. We
believe these recommendations will play an important role towards improving
repeatability in fields such as functional magnetic resonance imaging,
genomics, pharmacology, and more.
| stat |
The conditional permutation test for independence while controlling for
confounders | We propose a general new method, the conditional permutation test, for
testing the conditional independence of variables $X$ and $Y$ given a
potentially high-dimensional random vector $Z$ that may contain confounding
factors. The proposed test permutes entries of $X$ non-uniformly, so as to
respect the existing dependence between $X$ and $Z$ and thus account for the
presence of these confounders. Like the conditional randomization test of
Cand\`es et al. (2018), our test relies on the availability of an approximation
to the distribution of $X \mid Z$. While Cand\`es et al. (2018)'s test uses
this estimate to draw new $X$ values, for our test we use this approximation to
design an appropriate non-uniform distribution on permutations of the $X$
values already seen in the true data. We provide an efficient Markov Chain
Monte Carlo sampler for the implementation of our method, and establish bounds
on the Type I error in terms of the error in the approximation of the
conditional distribution of $X\mid Z$, finding that, for the worst case test
statistic, the inflation in Type I error of the conditional permutation test is
no larger than that of the conditional randomization test. We validate these
theoretical results with experiments on simulated data and on the Capital
Bikeshare data set.
| stat |
Shallow Self-Learning for Reject Inference in Credit Scoring | Credit scoring models support loan approval decisions in the financial
services industry. Lenders train these models on data from previously granted
credit applications, where the borrowers' repayment behavior has been observed.
This approach creates sample bias. The scoring model (i.e., classifier) is
trained on accepted cases only. Applying the resulting model to screen credit
applications from the population of all borrowers degrades model performance.
Reject inference comprises techniques to overcome sampling bias through
assigning labels to rejected cases. The paper makes two contributions. First,
we propose a self-learning framework for reject inference. The framework is
geared toward real-world credit scoring requirements through considering
distinct training regimes for iterative labeling and model training. Second, we
introduce a new measure to assess the effectiveness of reject inference
strategies. Our measure leverages domain knowledge to avoid artificial labeling
of rejected cases during strategy evaluation. We demonstrate this approach to
offer a robust and operational assessment of reject inference strategies.
Experiments on a real-world credit scoring data set confirm the superiority of
the adjusted self-learning framework over regular self-learning and previous
reject inference strategies. We also find strong evidence in favor of the
proposed evaluation measure assessing reject inference strategies more
reliably, raising the performance of the eventual credit scoring model.
| stat |
Generative Synthesis of Insurance Datasets | One of the impediments in advancing actuarial research and developing open
source assets for insurance analytics is the lack of realistic publicly
available datasets. In this work, we develop a workflow for synthesizing
insurance datasets leveraging CTGAN, a recently proposed neural network
architecture for generating tabular data. Applying the proposed workflow to
publicly available data in the domains of general insurance pricing and life
insurance shock lapse modeling, we evaluate the synthesized datasets from a few
perspectives: machine learning efficacy, distributions of variables, and
stability of model parameters. This workflow is implemented via an R interface
to promote adoption by researchers and data owners.
| stat |
Signed Sequential Rank Shiryaev-Roberts Schemes | We develop Shiryaev-Roberts schemes based on signed sequential ranks to
detect a persistent change in location of a continuous symmetric distribution
with known median. The in-control properties of these schemes are distribution
free, hence they do not require a parametric specification of an underlying
density function or the existence of any moments. Tables of control limits are
provided. The out-of-control average run length properties of the schemes are
gauged via theory-based calculations and Monte Carlo simulation. Comparisons
are made with two existing distribution-free schemes. We conclude that the
newly proposed scheme has much to recommend its use in practice. Implementation
of the methodology is illustrated in an application to a data set from an
industrial environment.
| stat |
Scalable approximate inference for state space models with normalising
flows | By exploiting mini-batch stochastic gradient optimisation, variational
inference has had great success in scaling up approximate Bayesian inference to
big data. To date, however, this strategy has only been applicable to models of
independent data. Here we extend mini-batch variational methods to state space
models of time series data. To do so we introduce a novel generative model as
our variational approximation, a local inverse autoregressive flow. This allows
a subsequence to be sampled without sampling the entire distribution. Hence we
can perform training iterations using short portions of the time series at low
computational cost. We illustrate our method on AR(1), Lotka-Volterra and
FitzHugh-Nagumo models, achieving accurate parameter estimation in a short
time.
| stat |
A Decomposition Analysis of Diffusion over a Large Network | Diffusion over a network refers to the phenomenon of a change of state of a
cross-sectional unit in one period leading to a change of state of its
neighbors in the network in the next period. One may estimate or test for
diffusion by estimating a cross-sectionally aggregated correlation between
neighbors over time from data. However, the estimated diffusion can be
misleading if the diffusion is confounded by omitted covariates. This paper
focuses on the measure of diffusion proposed by He and Song (2021), provides a
method of decomposition analysis to measure the role of the covariates on the
estimated diffusion, and develops an asymptotic inference procedure for the
decomposition analysis in such a situation. This paper also presents results
from a Monte Carlo study on the small sample performance of the inference
procedure.
| stat |
Recurrent Events Analysis With Data Collected at Informative Clinical
Visits in Electronic Health Records | Although increasingly used as a data resource for assembling cohorts,
electronic health records (EHRs) pose many analytic challenges. In particular,
a patient's health status influences when and what data are recorded,
generating sampling bias in the collected data. In this paper, we consider
recurrent event analysis using EHR data. Conventional regression methods for
event risk analysis usually require the values of covariates to be observed
throughout the follow-up period. In EHR databases, time-dependent covariates
are intermittently measured during clinical visits, and the timing of these
visits is informative in the sense that it depends on the disease course.
Simple methods, such as the last-observation-carried-forward approach, can lead
to biased estimation. On the other hand, complex joint models require
additional assumptions on the covariate process and cannot be easily extended
to handle multiple longitudinal predictors. By incorporating sampling weights
derived from estimating the observation time process, we develop a novel
estimation procedure based on inverse-rate-weighting and kernel-smoothing for
the semiparametric proportional rate model of recurrent events. The proposed
methods do not require model specifications for the covariate processes and can
easily handle multiple time-dependent covariates. Our methods are applied to a
kidney transplant study for illustration.
| stat |
Prediction in locally stationary time series | We develop an estimator for the high-dimensional covariance matrix of a
locally stationary process with a smoothly varying trend and use this statistic
to derive consistent predictors in non-stationary time series. In contrast to
the currently available methods for this problem the predictor developed here
does not rely on fitting an autoregressive model and does not require a
vanishing trend. The finite sample properties of the new methodology are
illustrated by means of a simulation study and a financial indices study.
| stat |
Robust Aggregation for Federated Learning | We present a robust aggregation approach to make federated learning robust to
settings when a fraction of the devices may be sending corrupted updates to the
server. The proposed approach relies on a robust secure aggregation oracle
based on the geometric median, which returns a robust aggregate using a
constant number of calls to a regular non-robust secure average oracle. The
robust aggregation oracle is privacy-preserving, similar to the secure average
oracle it builds upon. We provide experimental results of the proposed approach
with linear models and deep networks for two tasks in computer vision and
natural language processing. The robust aggregation approach is agnostic to the
level of corruption; it outperforms the classical aggregation approach in terms
of robustness when the level of corruption is high, while being competitive in
the regime of low corruption.
| stat |
Semi-supervised deep learning for high-dimensional uncertainty
quantification | Conventional uncertainty quantification methods usually lacks the capability
of dealing with high-dimensional problems due to the curse of dimensionality.
This paper presents a semi-supervised learning framework for dimension
reduction and reliability analysis. An autoencoder is first adopted for mapping
the high-dimensional space into a low-dimensional latent space, which contains
a distinguishable failure surface. Then a deep feedforward neural network (DFN)
is utilized to learn the mapping relationship and reconstruct the latent space,
while the Gaussian process (GP) modeling technique is used to build the
surrogate model of the transformed limit state function. During the training
process of the DFN, the discrepancy between the actual and reconstructed latent
space is minimized through semi-supervised learning for ensuring the accuracy.
Both labeled and unlabeled samples are utilized for defining the loss function
of the DFN. Evolutionary algorithm is adopted to train the DFN, then the Monte
Carlo simulation method is used for uncertainty quantification and reliability
analysis based on the proposed framework. The effectiveness is demonstrated
through a mathematical example.
| stat |
Discussion of the paper "Optimal treatment allocations in space and time
for on-line control of an emerging infectious disease'' by E. B. Laber, N. J.
Meyer, B. J. Reich, K. Pacifici, J. A. Collazo and J. Drake | This is a written discussion of the paper "Optimal treatment allocations in
space and time for on-line control of an emerging infectious disease" by E. B.
Laber, N. J. Meyer, B. J. Reich, K. Pacifici, J. A. Collazo and J. Drake,
contributed to the Journal of the Royal Statistical Society Series C
| stat |
An ensemble learning method for variable selection: application to high
dimensional data and missing values | Standard approaches for variable selection in linear models are not tailored
to deal properly with high dimensional and incomplete data. Currently, methods
dedicated to high dimensional data handle missing values by ad-hoc strategies,
like complete case analysis or single imputation, while methods dedicated to
missing values, mainly based on multiple imputation, do not discuss the
imputation method to use with high dimensional data. Consequently, both
approaches appear to be limited for many modern applications.
With inspiration from ensemble methods, a new variable selection method is
proposed. It extends classical variable selection methods such as stepwise,
lasso or knockoff in the case of high dimensional data with or without missing
data. Theoretical properties are studied and the practical interest is
demonstrated through a simulation study.
In the low dimensional case, the procedure improves the control of the error
risks, especially type I error, even without missing values. With missing
values, the method performs better than reference selection methods based on
multiple imputation. Similar performances are obtained in the high-dimensional
case with or without missing values.
| stat |
DNNSurv: Deep Neural Networks for Survival Analysis Using Pseudo Values | There has been increasing interest in modelling survival data using deep
learning methods in medical research. Current approaches have focused on
designing special cost functions to handle censored survival data. We propose a
very different method with two steps. In the first step, we transform each
subject's survival time into a series of jackknife pseudo conditional survival
probabilities and then use these pseudo probabilities as a quantitative
response variable in the deep neural network model. By using the pseudo values,
we reduce a complex survival analysis to a standard regression problem, which
greatly simplifies the neural network construction. Our two-step approach is
simple, yet very flexible in making risk predictions for survival data, which
is very appealing from the practice point of view. The source code is freely
available at http://github.com/lilizhaoUM/DNNSurv.
| stat |
Estimating Real Log Canonical Thresholds | Evaluation of the marginal likelihood plays an important role in model
selection problems. The widely applicable Bayesian information criterion (WBIC)
and singular Bayesian information criterion (sBIC) give approximations to the
log marginal likelihood, which can be applied to both regular and singular
models. When the real log canonical thresholds are known, the performance of
sBIC is considered to be better than that of WBIC, but only few real log
canonical thresholds are known. In this paper, we propose a new estimator of
the real log canonical thresholds based on the variance of thermodynamic
integration with an inverse temperature. In addition, we propose an application
to make sBIC widely applicable. Finally, we investigate the performance of the
estimator and model selection by simulation studies and application to real
data.
| stat |
Forecasting intra-individual changes of affective states taking into
account inter-individual differences using intensive longitudinal data from a
university student drop out study in math | The longitudinal process that leads to university student drop out in STEM
subjects can be described by referring to a) inter-individual differences
(e.g., cognitive abilities) as well as b) intra-individual changes (e.g.,
affective states), c) (unobserved) heterogeneity of trajectories, and d)
time-dependent variables. Large dynamic latent variable model frameworks for
intensive longitudinal data (ILD) have been proposed which are (partially)
capable of simultaneously separating the complex data structures (e.g., DLCA;
Asparouhov, Hamaker, & Muth\'en, 2017; DSEM; Asparouhov, Hamaker, & Muth\'en,
2018; NDLC-SEM, Kelava & Brandt, 2019). From a methodological perspective,
forecasting in dynamic frameworks allowing for real-time inferences on latent
or observed variables based on ongoing data collection has not been an
extensive research topic. From a practical perspective, there has been no
empirical study on student drop out in math that integrates ILD, dynamic
frameworks, and forecasting of critical states of the individuals allowing for
real-time interventions. In this paper, we show how Bayesian forecasting of
multivariate intra-individual variables and time-dependent class membership of
individuals (affective states) can be performed in these dynamic frameworks. To
illustrate our approach, we use an empirical example where we apply forecasting
methodology to ILD from a large university student drop out study in math with
multivariate observations collected over 50 measurement occasions from multiple
students (N = 122). More specifically, we forecast emotions and behavior
related to drop out. This allows us to model (i) just-in-time interventions,
(ii) detection of heterogeneity in trajectories, and (iii) prediction of
emerging dynamic states (e.g. critical stress levels or pre-decisional states).
| stat |
A Bayesian binary algorithm for RMS-based acoustic signal segmentation | Changepoint analysis (also known as segmentation analysis) aims at analyzing
an ordered, one-dimensional vector, in order to find locations where some
characteristic of the data changes. Many models and algorithms have been
studied under this theme, including models for changes in mean and / or
variance, changes in linear regression parameters, etc. In this work, we are
interested in an algorithm for the segmentation of long duration acoustic
signals; the segmentation is based on the change of the RMS power of the
signal. We investigate a Bayesian model with two possible parameterizations,
and propose a binary algorithm in two versions, using non-informative or
informative priors. We apply our algorithm to the segmentation of annotated
acoustic signals from the Alcatrazes marine preservation park in Brazil.
| stat |
A Markov Chain Monte-Carlo Approach to Dose-Response Optimization Using
Probabilistic Programming (RStan) | A hierarchical logistic regression Bayesian model is proposed and implemented
in R to model the probability of patient improvement corresponding to any given
dosage of a certain drug. RStan is used to obtain samples from the posterior
distributions via Markov Chain Monte-Carlo (MCMC) sampling. The effects of
selecting different families of prior distributions are examined and finally,
the posterior distributions are compared across RStan, and two other
environments, namely PyMC, and AgenaRisk.
| stat |
Nonnegative Bayesian nonparametric factor models with completely random
measures for community detection | We present a Bayesian nonparametric Poisson factorization model for modeling
network data with an unknown and potentially growing number of overlapping
communities. The construction is based on completely random measures and allows
the number of communities to either increase with the number of nodes at a
specified logarithmic or polynomial rate, or be bounded. We develop asymptotics
for the number of nodes and the degree distribution of the network and derive a
Markov chain Monte Carlo algorithm for targeting the exact posterior
distribution for this model. The usefulness of the approach is illustrated on
various real networks.
| stat |
Practical Bayesian System Identification using Hamiltonian Monte Carlo | This paper addresses Bayesian system identification using a Markov Chain
Monte Carlo approach. In particular, the Metroplis-Hastings algorithm with a
Hamiltonian proposal - known as Hamiltonian Monte Carlo - is reviewed and
adapted to linear and nonlinear system identification problems. The paper
details how the Hamiltonian proposal can be arranged so that the associated
Markov Chain performs well within the Metropolis-Hastings setting, which is a
key practical challenge faced when using the latter approach for many system
identification problems. This combination is demonstrated on several examples,
ranging from simple linear to more complex nonlinear systems, on both simulated
and real data.
| stat |
Robust Unsupervised Domain Adaptation for Neural Networks via Moment
Alignment | A novel approach for unsupervised domain adaptation for neural networks is
proposed. It relies on metric-based regularization of the learning process. The
metric-based regularization aims at domain-invariant latent feature
representations by means of maximizing the similarity between domain-specific
activation distributions. The proposed metric results from modifying an
integral probability metric such that it becomes less translation-sensitive on
a polynomial function space. The metric has an intuitive interpretation in the
dual space as the sum of differences of higher order central moments of the
corresponding activation distributions. Under appropriate assumptions on the
input distributions, error minimization is proven for the continuous case. As
demonstrated by an analysis of standard benchmark experiments for sentiment
analysis, object recognition and digit recognition, the outlined approach is
robust regarding parameter changes and achieves higher classification
accuracies than comparable approaches. The source code is available at
https://github.com/wzell/mann.
| stat |
Efficient Long-Range Convolutions for Point Clouds | The efficient treatment of long-range interactions for point clouds is a
challenging problem in many scientific machine learning applications. To
extract global information, one usually needs a large window size, a large
number of layers, and/or a large number of channels. This can often
significantly increase the computational cost. In this work, we present a novel
neural network layer that directly incorporates long-range information for a
point cloud. This layer, dubbed the long-range convolutional (LRC)-layer,
leverages the convolutional theorem coupled with the non-uniform Fourier
transform. In a nutshell, the LRC-layer mollifies the point cloud to an
adequately sized regular grid, computes its Fourier transform, multiplies the
result by a set of trainable Fourier multipliers, computes the inverse Fourier
transform, and finally interpolates the result back to the point cloud. The
resulting global all-to-all convolution operation can be performed in
nearly-linear time asymptotically with respect to the number of input points.
The LRC-layer is a particularly powerful tool when combined with local
convolution as together they offer efficient and seamless treatment of both
short and long range interactions. We showcase this framework by introducing a
neural network architecture that combines LRC-layers with short-range
convolutional layers to accurately learn the energy and force associated with a
$N$-body potential. We also exploit the induced two-level decomposition and
propose an efficient strategy to train the combined architecture with a reduced
number of samples.
| stat |
A Chi-square Goodness-of-Fit Test for Continuous Distributions against a
known Alternative | The chi square goodness-of-fit test is among the oldest known statistical
tests, first proposed by Pearson in 1900 for the multinomial distribution. It
has been in use in many fields ever since. However, various studies have shown
that when applied to data from a continuous distribution it is generally
inferior to other methods such as the Kolmogorov-Smirnov or Anderson-Darling
tests. However, the performance, that is the power, of the chi square test
depends crucially on the way the data is binned. In this paper we describe a
method that automatically finds a binning that is very good against a specific
alternative. We show that then the chi square test is generally competitive and
sometimes even superior to other standard tests.
| stat |
Diversity-Preserving K-Armed Bandits, Revisited | We consider the bandit-based framework for diversity-preserving
recommendations introduced by Celis et al. (2019), who approached it mainly by
a reduction to the setting of linear bandits. We design a UCB algorithm using
the specific structure of the setting and show that it enjoys a bounded
distribution-dependent regret in the natural cases when the optimal mixed
actions put some probability mass on all actions (i.e., when diversity is
desirable). Simulations illustrate this fact. We also provide regret lower
bounds and briefly discuss distribution-free regret bounds.
| stat |
Deep Learning with Functional Inputs | We present a methodology for integrating functional data into deep densely
connected feed-forward neural networks. The model is defined for scalar
responses with multiple functional and scalar covariates. A by-product of the
method is a set of dynamic functional weights that can be visualized during the
optimization process. This visualization leads to greater interpretability of
the relationship between the covariates and the response relative to
conventional neural networks. The model is shown to perform well in a number of
contexts including prediction of new data and recovery of the true underlying
functional weights; these results were confirmed through real applications and
simulation studies. A forthcoming R package is developed on top of a popular
deep learning library (Keras) allowing for general use of the approach.
| stat |
Inferring Hierarchical Mixture Structures:A Bayesian Nonparametric
Approach | This paper focuses on the problem of hierarchical non-overlapping clustering
of a dataset. In such a clustering, each data item is associated with exactly
one leaf node and each internal node is associated with all the data items
stored in the sub-tree beneath it, so that each level of the hierarchy
corresponds to a partition of the dataset. We develop a novel Bayesian
nonparametric method combining the nested Chinese Restaurant Process (nCRP) and
the Hierarchical Dirichlet Process (HDP). Compared with other existing Bayesian
approaches, our solution tackles data with complex latent mixture features
which has not been previously explored in the literature. We discuss the
details of the model and the inference procedure. Furthermore, experiments on
three datasets show that our method achieves solid empirical results in
comparison with existing algorithms.
| stat |
Random Feature Stein Discrepancies | Computable Stein discrepancies have been deployed for a variety of
applications, ranging from sampler selection in posterior inference to
approximate Bayesian inference to goodness-of-fit testing. Existing
convergence-determining Stein discrepancies admit strong theoretical guarantees
but suffer from a computational cost that grows quadratically in the sample
size. While linear-time Stein discrepancies have been proposed for
goodness-of-fit testing, they exhibit avoidable degradations in testing
power---even when power is explicitly optimized. To address these shortcomings,
we introduce feature Stein discrepancies ($\Phi$SDs), a new family of quality
measures that can be cheaply approximated using importance sampling. We show
how to construct $\Phi$SDs that provably determine the convergence of a sample
to its target and develop high-accuracy approximations---random $\Phi$SDs
(R$\Phi$SDs)---which are computable in near-linear time. In our experiments
with sampler selection for approximate posterior inference and goodness-of-fit
testing, R$\Phi$SDs perform as well or better than quadratic-time KSDs while
being orders of magnitude faster to compute.
| stat |
The Evolution of Dynamic Gaussian Process Model with Applications to
Malaria Vaccine Coverage Prediction | Gaussian process (GP) based statistical surrogates are popular, inexpensive
substitutes for emulating the outputs of expensive computer models that
simulate real-world phenomena or complex systems. Here, we discuss the
evolution of dynamic GP model - a computationally efficient statistical
surrogate for a computer simulator with time series outputs. The main idea is
to use a convolution of standard GP models, where the weights are guided by a
singular value decomposition (SVD) of the response matrix over the time
component. The dynamic GP model also adopts a localized modeling approach for
building a statistical model for large datasets.
In this chapter, we use several popular test function based computer
simulators to illustrate the evolution of dynamic GP models. We also use this
model for predicting the coverage of Malaria vaccine worldwide. Malaria is
still affecting more than eighty countries concentrated in the tropical belt.
In 2019 alone, it was the cause of more than 435,000 deaths worldwide. The
malice is easy to cure if diagnosed in time, but the common symptoms make it
difficult. We focus on a recently discovered reliable vaccine called Mos-Quirix
(RTS,S) which is currently going under human trials. With the help of publicly
available data on dosages, efficacy, disease incidence and communicability of
other vaccines obtained from the World Health Organisation, we predict vaccine
coverage for 78 Malaria-prone countries.
| stat |
ASAP: Architecture Search, Anneal and Prune | Automatic methods for Neural Architecture Search (NAS) have been shown to
produce state-of-the-art network models. Yet, their main drawback is the
computational complexity of the search process. As some primal methods
optimized over a discrete search space, thousands of days of GPU were required
for convergence. A recent approach is based on constructing a differentiable
search space that enables gradient-based optimization, which reduces the search
time to a few days. While successful, it still includes some noncontinuous
steps, e.g., the pruning of many weak connections at once. In this paper, we
propose a differentiable search space that allows the annealing of architecture
weights, while gradually pruning inferior operations. In this way, the search
converges to a single output network in a continuous manner. Experiments on
several vision datasets demonstrate the effectiveness of our method with
respect to the search cost and accuracy of the achieved model. Specifically,
with $0.2$ GPU search days we achieve an error rate of $1.68\%$ on CIFAR-10.
| stat |
What is the value of experimentation & measurement? | Experimentation and Measurement (E&M) capabilities allow organizations to
accurately assess the impact of new propositions and to experiment with many
variants of existing products. However, until now, the question of measuring
the measurer, or valuing the contribution of an E&M capability to
organizational success has not been addressed. We tackle this problem by
analyzing how, by decreasing estimation uncertainty, E&M platforms allow for
better prioritization. We quantify this benefit in terms of expected relative
improvement in the performance of all new propositions and provide guidance for
how much an E&M capability is worth and when organizations should invest in
one.
| stat |
Robust non-parametric mortality and fertility modelling and forecasting:
Gaussian process regression approaches | A rapid decline in mortality and fertility has become major issues in many
developed countries over the past few decades. A precise model for forecasting
demographic movements is important for decision making in social welfare
policies and resource budgeting among the government and many industry sectors.
This article introduces a novel non-parametric approach using Gaussian process
regression with a natural cubic spline mean function and a spectral mixture
covariance function for mortality and fertility modelling and forecasting.
Unlike most of the existing approaches in demographic modelling literature,
which rely on time parameters to decide the movements of the whole mortality or
fertility curve shifting from one year to another over time, we consider the
mortality and fertility curves from their components of all age-specific
mortality and fertility rates and assume each of them following a Gaussian
process over time to fit the whole curves in a discrete but intensive style.
The proposed Gaussian process regression approach shows significant
improvements in terms of preciseness and robustness compared to other
mainstream demographic modelling approaches in the short-, mid- and long-term
forecasting using the mortality and fertility data of several developed
countries in our numerical experiments.
| stat |
StaTIX - Statistical Type Inference on Linked Data | Large knowledge bases typically contain data adhering to various schemas with
incomplete and/or noisy type information. This seriously complicates further
integration and post-processing efforts, as type information is crucial in
correctly handling the data. In this paper, we introduce a novel statistical
type inference method, called StaTIX, to effectively infer instance types in
Linked Data sets in a fully unsupervised manner. Our inference technique
leverages a new hierarchical clustering algorithm that is robust, highly
effective, and scalable. We introduce a novel approach to reduce the processing
complexity of the similarity matrix specifying the relations between various
instances in the knowledge base. This approach speeds up the inference process
while also improving the correctness of the inferred types due to the noise
attenuation in the input data. We further optimize the clustering process by
introducing a dedicated hash function that speeds up the inference process by
orders of magnitude without negatively affecting its accuracy. Finally, we
describe a new technique to identify representative clusters from the
multi-scale output of our clustering algorithm to further improve the accuracy
of the inferred types. We empirically evaluate our approach on several
real-world datasets and compare it to the state of the art. Our results show
that StaTIX is more efficient than existing methods (both in terms of speed and
memory consumption) as well as more effective. StaTIX reduces the F1-score
error of the predicted types by about 40% on average compared to the state of
the art and improves the execution time by orders of magnitude.
| stat |
Unified Principal Component Analysis for Sparse and Dense Functional
Data under Spatial Dependency | We consider spatially dependent functional data collected under a
geostatistics setting, where spatial locations are irregular and random. The
functional response is the sum of a spatially dependent functional effect and a
spatially independent functional nugget effect. Observations on each function
are made on discrete time points and contaminated with measurement errors.
Under the assumption of spatial stationarity and isotropy, we propose a tensor
product spline estimator for the spatio-temporal covariance function. When a
coregionalization covariance structure is further assumed, we propose a new
functional principal component analysis method that borrows information from
neighboring functions. The proposed method also generates nonparametric
estimators for the spatial covariance functions, which can be used for
functional kriging. Under a unified framework for sparse and dense functional
data, infill and increasing domain asymptotic paradigms, we develop the
asymptotic convergence rates for the proposed estimators. Advantages of the
proposed approach are demonstrated through simulation studies and two real data
applications representing sparse and dense functional data, respectively.
| stat |
Quantification of predictive uncertainty in hydrological modelling by
harnessing the wisdom of the crowd: A large-sample experiment at monthly
timescale | Predictive hydrological uncertainty can be quantified by using ensemble
methods. If properly formulated, these methods can offer improved predictive
performance by combining multiple predictions. In this work, we use
50-year-long monthly time series observed in 270 catchments in the United
States to explore the performances provided by an ensemble learning
post-processing methodology for issuing probabilistic hydrological predictions.
This methodology allows the utilization of flexible quantile regression models
for exploiting information about the hydrological model's error. Its key
differences with respect to basic two-stage hydrological post-processing
methodologies using the same type of regression models are that (a) instead of
a single point hydrological prediction it generates a large number of "sister
predictions" (yet using a single hydrological model), and that (b) it relies on
the concept of combining probabilistic predictions via simple quantile
averaging. A major hydrological modelling challenge is obtaining probabilistic
predictions that are simultaneously reliable and associated to prediction bands
that are as narrow as possible; therefore, we assess both these desired
properties of the predictions by computing their coverage probabilities,
average widths and average interval scores. The results confirm the usefulness
of the proposed methodology and its larger robustness with respect to basic
two-stage post-processing methodologies. Finally, this methodology is
empirically proven to harness the "wisdom of the crowd" in terms of average
interval score, i.e., the average of the individual predictions combined by
this methodology scores no worse -- usually better -- than the average of the
scores of the individual predictions.
| stat |
Subsets and Splits
No saved queries yet
Save your SQL queries to embed, download, and access them later. Queries will appear here once saved.