title
stringlengths 6
244
| abstract
stringlengths 19
5.09k
| label
stringclasses 10
values |
|---|---|---|
CDPA: Common and Distinctive Pattern Analysis between High-dimensional
Datasets | A representative model in integrative analysis of two high-dimensional
correlated datasets is to decompose each data matrix into a low-rank common
matrix generated by latent factors shared across datasets, a low-rank
distinctive matrix corresponding to each dataset, and an additive noise matrix.
Existing decomposition methods claim that their common matrices capture the
common pattern of the two datasets. However, their so-called common pattern
only denotes the common latent factors but ignores the common information
between the two coefficient matrices of these latent factors. We propose a
novel method, called the common and distinctive pattern analysis (CDPA), which
appropriately defines the two patterns by further incorporating the common and
distinctive information of the coefficient matrices. A consistent estimation
approach is developed for high-dimensional settings, and shows reasonably good
finite-sample performance in simulations. The superiority of CDPA over
state-of-the-art methods is corroborated in both simulated data and two
real-data examples from the Human Connectome Project and The Cancer Genome
Atlas. A Python package implementing the CDPA method is available at
https://github.com/shu-hai/CDPA.
| stat |
Fundamental Limits of Deep Graph Convolutional Networks | Graph convolutional networks (GCNs) are a widely used method for graph
representation learning. To elucidate the capabilities and limitations of GCNs,
we investigate their power, as a function of their number of layers, to
distinguish between different random graph models (corresponding to different
class-conditional distributions in a classification problem) on the basis of
the embeddings of their sample graphs. In particular, the graph models that we
consider arise from graphons, which are the most general possible
parameterizations of infinite exchangeable graph models and which are the
central objects of study in the theory of dense graph limits. We give a precise
characterization of the set of pairs of graphons that are indistinguishable by
a GCN with nonlinear activation functions coming from a certain broad class if
its depth is at least logarithmic in the size of the sample graph. This
characterization is in terms of a degree profile closeness property. Outside
this class, a very simple GCN architecture suffices for distinguishability. We
then exhibit a concrete, infinite class of graphons arising from stochastic
block models that are well-separated in terms of cut distance and are
indistinguishable by a GCN. These results theoretically match empirical
observations of several prior works. To prove our results, we exploit a
connection to random walks on graphs. Finally, we give empirical results on
synthetic and real graph classification datasets, indicating that
indistinguishable graph distributions arise in practice.
| stat |
qgam: Bayesian non-parametric quantile regression modelling in R | Generalized additive models (GAMs) are flexible non-linear regression models,
which can be fitted efficiently using the approximate Bayesian methods provided
by the mgcv R package. While the GAM methods provided by mgcv are based on the
assumption that the response distribution is modelled parametrically, here we
discuss more flexible methods that do not entail any parametric assumption. In
particular, this article introduces the qgam package, which is an extension of
mgcv providing fast calibrated Bayesian methods for fitting quantile GAMs
(QGAMs) in R. QGAMs are based on a smooth version of the pinball loss of
Koenker (2005), rather than on a likelihood function, hence jointly achieving
satisfactory accuracy of the quantile point estimates and coverage of the
corresponding credible intervals requires adopting the specialized Bayesian
fitting framework of Fasiolo, Wood, Zaffran, Nedellec, and Goude (2020b). Here
we detail how this framework is implemented in qgam and we provide examples
illustrating how the package should be used in practice.
| stat |
Detecting multiple generalized change-points by isolating single ones | We introduce a new approach, called Isolate-Detect (ID), for the consistent
estimation of the number and location of multiple generalized change-points in
noisy data sequences. Examples of signal changes that ID can deal with, are
changes in the mean of a piecewise-constant signal and changes in the trend,
accompanied by discontinuities or not, in the piecewise-linear model. The
number of change-points can increase with the sample size. Our method is based
on an isolation technique, which prevents the consideration of intervals that
contain more than one change-point. This isolation enhances ID's accuracy as it
allows for detection in the presence of frequent changes of possibly small
magnitudes. Thresholding and model selection through an information criterion
are the two stopping rules described in the article. A hybrid of both criteria
leads to a general method with very good practical performance and minimal
parameter choice. In the scenarios tested, ID is at least as accurate as the
state-of-the-art methods; most of the times it outperforms them. The R package
IDetect implementing the method from the paper is available from CRAN.
| stat |
A calibration framework for high-resolution hydrological models using a
multiresolution and heterogeneous strategy | Increasing spatial and temporal resolution of numerical models continues to
propel progress in hydrological sciences, but, at the same time, it has
strained the ability of modern automatic calibration methods to produce
realistic model parameter combinations for these models. This paper presents a
new reliable and fast automatic calibration framework to address this issue. In
essence, the proposed framework, adopting a divide and conquer strategy, first
partitions the parameters into groups of different resolutions based on their
sensitivity or importance, in which the most sensitive parameters are
prioritized with highest resolution in parameter search space, while the least
sensitive ones are explored with the coarsest resolution at beginning. This is
followed by an optimization based iterative calibration procedure consisting of
a series of sub-tasks or runs. Between consecutive runs, the setup
configuration is heterogeneous with parameter search ranges and resolutions
varying among groups. At the completion of each sub-task, the parameter ranges
within each group are systematically refined from their previously estimated
ranges which are initially based on a priori information. Parameters attain
stable convergence progressively with each run. A comparison of this new
calibration framework with a traditional optimization-based approach was
performed using a quasi-synthetic double-model setup experiment to calibrate
134 parameters and two well-known distributed hydrological models: the Variable
Infiltration Capacity (VIC) model and the Distributed Hydrology Soil Vegetation
Model (DHSVM). The results demonstrate statistically that the proposed
framework can better mitigate equifinality problem, yields more realistic model
parameter estimates, and is computationally more efficient.
| stat |
Additive Bayesian Network Modelling with the R Package abn | The R package abn is designed to fit additive Bayesian models to
observational datasets. It contains routines to score Bayesian networks based
on Bayesian or information theoretic formulations of generalized linear models.
It is equipped with exact search and greedy search algorithms to select the
best network. It supports a possible blend of continuous, discrete and count
data and input of prior knowledge at a structural level. The Bayesian
implementation supports random effects to control for one-layer clustering. In
this paper, we give an overview of the methodology and illustrate the package's
functionalities using a veterinary dataset about respiratory diseases in
commercial swine production.
| stat |
The SPDE Approach to Mat\'ern Fields: Graph Representations | This paper investigates Gaussian Markov random field approximations to
nonstationary Gaussian fields using graph representations of stochastic partial
differential equations. We establish approximation error guarantees building on
the theory of spectral convergence of graph Laplacians. The proposed graph
representations provide a generalization of the Mat\'ern model to unstructured
point clouds, and facilitate inference and sampling using linear algebra
methods for sparse matrices. In addition, they bridge and unify several models
in Bayesian inverse problems, spatial statistics and graph-based machine
learning. We demonstrate through examples in these three disciplines that the
unity revealed by graph representations facilitates the exchange of ideas
across them.
| stat |
Predictive clustering | We show how to convert any clustering into a prediction set. This has the
effect of converting the clustering into a (possibly overlapping) union of
spheres or ellipsoids. The tuning parameters can be chosen to minimize the size
of the prediction set. When applied to k-means clustering, this method solves
several problems: the method tells us how to choose k, how to merge clusters
and how to replace the Voronoi partition with more natural shapes. We show that
the same reasoning can be applied to other clustering methods.
| stat |
Predictive Inference Is Free with the Jackknife+-after-Bootstrap | Ensemble learning is widely used in applications to make predictions in
complex decision problems---for example, averaging models fitted to a sequence
of samples bootstrapped from the available training data. While such methods
offer more accurate, stable, and robust predictions and model estimates, much
less is known about how to perform valid, assumption-lean inference on the
output of these types of procedures. In this paper, we propose the
jackknife+-after-bootstrap (J+aB), a procedure for constructing a predictive
interval, which uses only the available bootstrapped samples and their
corresponding fitted models, and is therefore "free" in terms of the cost of
model fitting. The J+aB offers a predictive coverage guarantee that holds with
no assumptions on the distribution of the data, the nature of the fitted model,
or the way in which the ensemble of models are aggregated---at worst, the
failure rate of the predictive interval is inflated by a factor of 2. Our
numerical experiments verify the coverage and accuracy of the resulting
predictive intervals on real data.
| stat |
Longitudinal modeling of age-dependent latent traits with generalized
additive latent and mixed models | We present generalized additive latent and mixed models (GALAMMs) for
analysis of clustered data with latent and observed variables depending
smoothly on observed variables. A profile likelihood algorithm is proposed, and
we derive asymptotic standard errors of both smooth and parametric terms. The
work was motivated by applications in cognitive neuroscience, and we show how
GALAMMs can successfully model the complex lifespan trajectory of latent
episodic memory, along with a discrepant trajectory of working memory, as well
as the effect of latent socioeconomic status on hippocampal development.
Simulation experiments suggest that model estimates are accurate even with
moderate sample sizes.
| stat |
Evaluation of a meta-analysis of air quality and heart attacks, a case
study | It is generally acknowledged that claims from observational studies often
fail to replicate. An exploratory study was undertaken to assess the
reliability of base studies used in meta-analysis of short-term air
quality-myocardial infarction risk and to judge the reliability of statistical
evidence from meta-analysis that uses data from observational studies. A highly
cited meta-analysis paper examining whether short-term air quality exposure
triggers myocardial infarction was evaluated as a case study. The paper
considered six air quality components - carbon monoxide, nitrogen dioxide,
sulfur dioxide, particulate matter 10 and 2.5 micrometers in diameter (PM10 and
PM2.5), and ozone. The number of possible questions and statistical models at
issue in each of 34 base papers used were estimated and p-value plots for each
of the air components were constructed to evaluate the effect heterogeneity of
p-values used from the base papers. Analysis search spaces (number of
statistical tests possible) in the base papers were large, median of 12,288,
interquartile range: 2,496 to 58,368, in comparison to actual statistical test
results presented. Statistical test results taken from the base papers may not
provide unbiased measures of effect for meta-analysis. Shapes of p-value plots
for the six air components were consistent with the possibility of analysis
manipulation to obtain small p-values in several base papers. Results suggest
the appearance of heterogeneous, researcher-generated p-values used in the
meta-analysis rather than unbiased evidence of real effects for air quality. We
conclude that this meta-analysis does not provide reliable evidence for an
association of air quality components with myocardial risk.
| stat |
On the Optimality of Perturbations in Stochastic and Adversarial
Multi-armed Bandit Problems | We investigate the optimality of perturbation based algorithms in the
stochastic and adversarial multi-armed bandit problems. For the stochastic
case, we provide a unified regret analysis for both sub-Weibull and bounded
perturbations when rewards are sub-Gaussian. Our bounds are instance optimal
for sub-Weibull perturbations with parameter 2 that also have a matching lower
tail bound, and all bounded support perturbations where there is sufficient
probability mass at the extremes of the support. For the adversarial setting,
we prove rigorous barriers against two natural solution approaches using tools
from discrete choice theory and extreme value theory. Our results suggest that
the optimal perturbation, if it exists, will be of Frechet-type.
| stat |
The Spectrum of Fisher Information of Deep Networks Achieving Dynamical
Isometry | The Fisher information matrix (FIM) is fundamental to understanding the
trainability of deep neural nets (DNN), since it describes the parameter
space's local metric. We investigate the spectral distribution of the
conditional FIM, which is the FIM given a single sample, by focusing on
fully-connected networks achieving dynamical isometry. Then, while dynamical
isometry is known to keep specific backpropagated signals independent of the
depth, we find that the parameter space's local metric linearly depends on the
depth even under the dynamical isometry. More precisely, we reveal that the
conditional FIM's spectrum concentrates around the maximum and the value grows
linearly as the depth increases. To examine the spectrum, considering random
initialization and the wide limit, we construct an algebraic methodology based
on the free probability theory. As a byproduct, we provide an analysis of the
solvable spectral distribution in two-hidden-layer cases. Lastly, experimental
results verify that the appropriate learning rate for the online training of
DNNs is in inverse proportional to depth, which is determined by the
conditional FIM's spectrum.
| stat |
Precise Statistical Analysis of Classification Accuracies for
Adversarial Training | Despite the wide empirical success of modern machine learning algorithms and
models in a multitude of applications, they are known to be highly susceptible
to seemingly small indiscernible perturbations to the input data known as
adversarial attacks. A variety of recent adversarial training procedures have
been proposed to remedy this issue. Despite the success of such procedures at
increasing accuracy on adversarially perturbed inputs or robust accuracy, these
techniques often reduce accuracy on natural unperturbed inputs or standard
accuracy. Complicating matters further the effect and trend of adversarial
training procedures on standard and robust accuracy is rather counter intuitive
and radically dependent on a variety of factors including the perceived form of
the perturbation during training, size/quality of data, model
overparameterization, etc. In this paper we focus on binary classification
problems where the data is generated according to the mixture of two Gaussians
with general anisotropic covariance matrices and derive a precise
characterization of the standard and robust accuracy for a class of minimax
adversarially trained models. We consider a general norm-based adversarial
model, where the adversary can add perturbations of bounded $\ell_p$ norm to
each input data, for an arbitrary $p\ge 1$. Our comprehensive analysis allows
us to theoretically explain several intriguing empirical phenomena and provide
a precise understanding of the role of different problem parameters on standard
and robust accuracies.
| stat |
Bayesian forecasting of electoral outcomes with new parties' competition | This paper proposed a methodology to forecast electoral outcomes using the
result of the combination of a fundamental model and a model-based aggregation
of polls. We propose a Bayesian hierarchical structure for the fundamental
model that synthesises data at the provincial, regional and national level. We
use a Bayesian strategy to combine the fundamental model with the information
coming for recent polls. This model can naturally be updated every time new
information, for instance a new poll, becomes available. This methodology is
well suited to deal with increasingly frequent situations in which new
political parties enter an electoral competition, although our approach is
general enough to accommodate any other electoral situation. We illustrate the
advantages of our method using the 2015 Spanish Congressional Election in which
two new parties ended up receiving 30\% of the votes. We compare the predictive
performance of our model versus alternative models. In general the predictions
of our model outperform the alternative specifications, including hybrid models
that combine fundamental and polls models. Our predictions are, in relative
terms, particularly accurate in predicting the seats obtained by each political
party.
| stat |
Additive Tree-Structured Covariance Function for Conditional Parameter
Spaces in Bayesian Optimization | Bayesian optimization (BO) is a sample-efficient global optimization
algorithm for black-box functions which are expensive to evaluate. Existing
literature on model based optimization in conditional parameter spaces are
usually built on trees. In this work, we generalize the additive assumption to
tree-structured functions and propose an additive tree-structured covariance
function, showing improved sample-efficiency, wider applicability and greater
flexibility. Furthermore, by incorporating the structure information of
parameter spaces and the additive assumption in the BO loop, we develop a
parallel algorithm to optimize the acquisition function and this optimization
can be performed in a low dimensional space. We demonstrate our method on an
optimization benchmark function, as well as on a neural network model
compression problem, and experimental results show our approach significantly
outperforms the current state of the art for conditional parameter optimization
including SMAC, TPE and Jenatton et al. (2017).
| stat |
Uniform regret bounds over $R^d$ for the sequential linear regression
problem with the square loss | We consider the setting of online linear regression for arbitrary
deterministic sequences, with the square loss. We are interested in the aim set
by Bartlett et al. (2015): obtain regret bounds that hold uniformly over all
competitor vectors. When the feature sequence is known at the beginning of the
game, they provided closed-form regret bounds of $2d B^2 \ln T +
\mathcal{O}_T(1)$, where $T$ is the number of rounds and $B$ is a bound on the
observations. Instead, we derive bounds with an optimal constant of $1$ in
front of the $d B^2 \ln T$ term. In the case of sequentially revealed features,
we also derive an asymptotic regret bound of $d B^2 \ln T$ for any individual
sequence of features and bounded observations. All our algorithms are variants
of the online non-linear ridge regression forecaster, either with a
data-dependent regularization or with almost no regularization.
| stat |
The i3+3 Design for Phase I Clinical Trials | Purpose: The 3+3 design has been shown to be less likely to achieve the
objectives of phase I dose-finding trials when compared with more advanced
model-based designs. One major criticism of the 3+3 design is that it is based
on simple rules, does not depend on statistical models for inference, and leads
to unsafe and unreliable operating characteristics. On the other hand, being
rule-based allows 3+3 to be easily understood and implemented in practice,
making it the first choice among clinicians. Is it possible to have a
rule-based design with great performance? Methods: We propose a new rule-based
design called i3+3, where the letter "i" represents the word "interval". The
i3+3 design is based on simple but more advanced rules that account for the
variabilities in the observed data. We compare the operating characteristics
for the proposed i3+3 design with other popular phase I designs by simulation.
Results: The i3+3 design is far superior than the 3+3 design in trial safety
and the ability to identify the true MTD. Compared with model-based phase I
designs, i3+3 also demonstrates comparable performances. In other words, the
i3+3 design possesses both the simplicity and transparency of the rule-based
approaches, and the superior operating characteristics seen in model-based
approaches. An online R Shiny tool (https://i3design.shinyapps.io/i3plus3/) is
provided to illustrate the i3+3 design, although in practice it requires no
software to design or conduct a dose-finding trial. Conclusion: The i3+3 design
could be a practice-altering method for the clinical community.
| stat |
Uncertainty Quantification in Ensembles of Honest Regression Trees using
Generalized Fiducial Inference | Due to their accuracies, methods based on ensembles of regression trees are a
popular approach for making predictions. Some common examples include Bayesian
additive regression trees, boosting and random forests. This paper focuses on
honest random forests, which add honesty to the original form of random forests
and are proved to have better statistical properties. The main contribution is
a new method that quantifies the uncertainties of the estimates and predictions
produced by honest random forests. The proposed method is based on the
generalized fiducial methodology, and provides a fiducial density function that
measures how likely each single honest tree is the true model. With such a
density function, estimates and predictions, as well as their
confidence/prediction intervals, can be obtained. The promising empirical
properties of the proposed method are demonstrated by numerical comparisons
with several state-of-the-art methods, and by applications to a few real data
sets. Lastly, the proposed method is theoretically backed up by a strong
asymptotic guarantee.
| stat |
Marginally-calibrated deep distributional regression | Deep neural network (DNN) regression models are widely used in applications
requiring state-of-the-art predictive accuracy. However, until recently there
has been little work on accurate uncertainty quantification for predictions
from such models. We add to this literature by outlining an approach to
constructing predictive distributions that are `marginally calibrated'. This is
where the long run average of the predictive distributions of the response
variable matches the observed empirical margin. Our approach considers a DNN
regression with a conditionally Gaussian prior for the final layer weights,
from which an implicit copula process on the feature space is extracted. This
copula process is combined with a non-parametrically estimated marginal
distribution for the response. The end result is a scalable distributional DNN
regression method with marginally calibrated predictions, and our work
complements existing methods for probability calibration. The approach is first
illustrated using two applications of dense layer feed-forward neural networks.
However, our main motivating applications are in likelihood-free inference,
where distributional deep regression is used to estimate marginal posterior
distributions. In two complex ecological time series examples we employ the
implicit copulas of convolutional networks, and show that marginal calibration
results in improved uncertainty quantification. Our approach also avoids the
need for manual specification of summary statistics, a requirement that is
burdensome for users and typical of competing likelihood-free inference
methods.
| stat |
Identifying latent classes with ordered categorical indicators | A Monte Carlo simulation was used to determine which assumptions for ordered
categorical data, continuity vs. discrete categories, most frequently
identifies the underlying factor structure when a response variable has five
ordered categories. The impact of infrequently endorsed response categories was
also examined, a condition that has not been fully explored. The typical method
for overcoming infrequently endorsed categories in applied research is to
collapse response options with adjacent categories resulting in less response
categories that are endorsed more frequently, but this approach may not
necessarily provide useful information. Response category endorsement issues
have been studied in Item Response Theory, but this issue has not been
addressed in classification analyses nor has fit measure performance been
examined under these conditions. We found that the performance of commonly used
fit statistics to identify the true number of latent class depends on the
whether continuity is assumed, sample size, and convergence. Fit statistics
performed best when the five response options are assumed to be categorical.
However, in situations with lower sample sizes and when convergence is an
issue, assuming continuity and using the adjusted Lo-Mendell-Rubin likelihood
ratio test may be useful.
| stat |
Optimal spectral shrinkage and PCA with heteroscedastic noise | This paper studies the related problems of prediction, covariance estimation,
and principal component analysis for the spiked covariance model with
heteroscedastic noise. We consider an estimator of the principal components
based on whitening the noise, and we derive optimal singular value and
eigenvalue shrinkers for use with these estimated principal components.
Underlying these methods are new asymptotic results for the high-dimensional
spiked model with heteroscedastic noise, and consistent estimators for the
relevant population parameters. We extend previous analysis on out-of-sample
prediction to the setting of predictors with whitening. We demonstrate certain
advantages of noise whitening. Specifically, we show that in a certain
asymptotic regime, optimal singular value shrinkage with whitening converges to
the best linear predictor, whereas without whitening it converges to a
suboptimal linear predictor. We prove that for generic signals, whitening
improves estimation of the principal components, and increases a natural
signal-to-noise ratio of the observations. We also show that for rank one
signals, our estimated principal components achieve the asymptotic minimax
rate.
| stat |
Machine learning-based clinical prediction modeling -- A practical guide
for clinicians | In the emerging era of big data, larger available clinical datasets and
computational advances have sparked a massive interest in machine
learning-based approaches. The number of manuscripts related to machine
learning or artificial intelligence has exponentially increased over the past
years. As analytical machine learning tools become readily available for
clinicians to use, the understanding of key concepts and the awareness of
analytical pitfalls are increasingly required for clinicians, investigators,
reviewers and editors, who even as experts in their clinical field, sometimes
find themselves insufficiently equipped to evaluate machine learning
methodologies. In the first section, we provide explanations on the general
principles of machine learning, as well as analytical steps required for
successful machine learning-based predictive modelling - which is the focus of
this series. In further sections, we review the importance of resampling,
overfitting and model generalizability as well as feature reduction and
selection (Part II), strategies for model evaluation, reporting and discussion
of common caveats and other points of significance (Part III), as well as offer
a practical guide to classification (Part IV) and regression modelling (Part
V), with a complete coding pipeline. Methodological rigor and clarity as well
as understanding of the underlying reasoning of the internal workings of a
machine learning approach are required, otherwise predictive applications
despite being strong analytical tools are not well accepted into the clinical
routine. Going forward, machine learning and artificial intelligence shape and
influence modern medicine across disciplines including the field of
neurosurgery.
| stat |
Descent-to-Delete: Gradient-Based Methods for Machine Unlearning | We study the data deletion problem for convex models. By leveraging
techniques from convex optimization and reservoir sampling, we give the first
data deletion algorithms that are able to handle an arbitrarily long sequence
of adversarial updates while promising both per-deletion run-time and
steady-state error that do not grow with the length of the update sequence. We
also introduce several new conceptual distinctions: for example, we can ask
that after a deletion, the entire state maintained by the optimization
algorithm is statistically indistinguishable from the state that would have
resulted had we retrained, or we can ask for the weaker condition that only the
observable output is statistically indistinguishable from the observable output
that would have resulted from retraining. We are able to give more efficient
deletion algorithms under this weaker deletion criterion.
| stat |
Fast Multi-resolution Segmentation for Nonstationary Hawkes Process
Using Cumulants | The stationarity is assumed in vanilla Hawkes process, which reduces the
model complexity but introduces a strong assumption. In this paper, we propose
a fast multi-resolution segmentation algorithm to capture the time-varying
characteristics of nonstationary Hawkes process. The proposed algorithm is
based on the first and second order cumulants. Except for the computation
efficiency, the algorithm can provide a hierarchical view of the segmentation
at different resolutions. We extensively investigate the impact of
hyperparameters on the performance of this algorithm. To ease the choice of one
hyperparameter, a refined Gaussian process based segmentation algorithm is also
proposed which proves to be robust. The proposed algorithm is applied to a real
vehicle collision dataset and the outcome shows some interesting hierarchical
dynamic time-varying characteristics.
| stat |
Support Estimation with Sampling Artifacts and Errors | The problem of estimating the support of a distribution is of great
importance in many areas of machine learning, computer science, physics and
biology. Most of the existing work in this domain has focused on settings that
assume perfectly accurate sampling approaches, which is seldom true in
practical data science. Here we introduce the first known approach to support
estimation in the presence of sampling artifacts and errors where each sample
is assumed to arise from a Poisson repeat channel which simultaneously captures
repetitions and deletions of samples. The proposed estimator is based on
regularized weighted Chebyshev approximations, with weights governed by
evaluations of so-called Touchard (Bell) polynomials. The supports in the
presence of sampling artifacts are calculated using discretized semi-infite
programming methods. The estimation approach is tested on synthetic and textual
data, as well as on GISAID data collected to address a new problem in
computational biology: mutational support estimation in genes of the SARS-Cov-2
virus. In the later setting, the Poisson channel captures the fact that many
individuals are tested multiple times for the presence of viral RNA, thereby
leading to repeated samples, while other individual's results are not recorded
due to test errors. For all experiments performed, we observed significant
improvements of our integrated methods compared to those obtained through
adequate modifications of state-of-the-art noiseless support estimation
methods.
| stat |
RobustSleepNet: Transfer learning for automated sleep staging at scale | Sleep disorder diagnosis relies on the analysis of polysomnography (PSG)
records. Sleep stages are systematically determined as a preliminary step of
this examination. In practice, sleep stage classification relies on the visual
inspection of 30-seconds epochs of polysomnography signals. Numerous automatic
approaches have been developed to replace this tedious and expensive task.
Although these methods demonstrated better performance than human sleep experts
on specific datasets, they remain largely unused in sleep clinics. The main
reason is that each sleep clinic uses a specific PSG montage that most
automatic approaches are unable to handle out-of-the-box. Moreover, even when
the PSG montage is compatible, publications have shown that automatic
approaches perform poorly on unseen data with different demographics. To
address these issues, we introduce RobustSleepNet, a deep learning model for
automatic sleep stage classification able to handle arbitrary PSG montages. We
trained and evaluated this model in a leave-one-out-dataset fashion on a large
corpus of 8 heterogeneous sleep staging datasets to make it robust to
demographic changes. When evaluated on an unseen dataset, RobustSleepNet
reaches 97% of the F1 of a model trained specifically on this dataset. We then
show that finetuning RobustSleepNet, using a part of the unseen dataset,
increase the F1 by 2% when compared to a model trained specifically for this
dataset. Hence, RobustSleepNet unlocks the possibility to perform high-quality
out-of-the-box automatic sleep staging with any clinical setup. It can also be
finetuned to reach a state-of-the-art level of performance on a specific
population.
| stat |
Topological Techniques in Model Selection | The LASSO is an attractive regularisation method for linear regression that
combines variable selection with an efficient computation procedure. This paper
is concerned with enhancing the performance of LASSO for square-free
hierarchical polynomial models when combining validation error with a measure
of model complexity. The measure of the complexity is the sum of Betti numbers
of the model which is seen as a simplicial complex, and we describe the model
in terms of components and cycles, borrowing from recent developments in
computational topology. We study and propose an algorithm which combines
statistical and topological criteria. This compound criterion would allow us to
deal with model selection problems in polynomial regression models containing
higher-order interactions. Simulation results demonstrate that the compound
criteria produce sparser models with lower prediction errors than the
estimators of several other statistical methods for higher order interaction
models.
| stat |
Augmented Minimax Linear Estimation | Many statistical estimands can expressed as continuous linear functionals of
a conditional expectation function. This includes the average treatment effect
under unconfoundedness and generalizations for continuous-valued and
personalized treatments. In this paper, we discuss a general approach to
estimating such quantities: we begin with a simple plug-in estimator based on
an estimate of the conditional expectation function, and then correct the
plug-in estimator by subtracting a minimax linear estimate of its error. We
show that our method is semiparametrically efficient under weak conditions and
observe promising performance on both real and simulated data.
| stat |
P-spline smoothed functional ICA of EEG data | We propose a novel functional data framework for artifact extraction and
removal to estimate brain electrical activity sources from EEG signals. Our
methodology is derived on the basis of event related potential (ERP) analysis,
and motivated by mapping adverse artifactual events caused by body movements
and physiological activity originated outside the brain. A functional
independent component analysis (FICA) based on the use of fourth moments is
conducted on the principal component expansion in terms of B-spline basis
functions. We extend this model setup by introducing a discrete roughness
penalty in the orthonormality constraint of the functional principal component
decomposition to later compute estimates of FICA. Compared to other ICA
algorithms, our method combines a regularization mechanism stemmed from the
principal eigendirections with a discrete penalization given by the $d$-order
difference operator. In this regard, it allows to naturally control
high-frequency remnants of neural origin overlapping latent artifactual
eigenfunctions and thus to preserve this persistent activity at artifact
extraction level. Furthermore, we introduce a new cross-validation method for
the selection of the penalization parameter which uses shrinkage to asses the
performance of the estimators for functional representations with larger basis
dimension and excess of roughness. This method is used in combination with a
kurtosis measure in order to provide the optimal number of independent
components.The FICA model is illustrated at functional and longitudinal
dimensions by an example on real EEG data where a subject willingly performs
arm gestures and stereotyped physiological artifacts. Our method can be
relevant in neurocognitive research and related fields, particularlly in
situations where movement can bias the estimation of brain potentials.
| stat |
Partial Identification of Nonseparable Models using Binary Instruments | In this study, we explore the partial identification of nonseparable models
with continuous endogenous and binary instrumental variables. We show that the
structural function is partially identified when it is monotone or concave in
the explanatory variable. D'Haultfoeuille and Fevrier (2015) and Torgovitsky
(2015) prove the point identification of the structural function under a key
assumption that the conditional distribution functions of the endogenous
variable for different values of the instrumental variables have intersections.
We demonstrate that, even if this assumption does not hold, monotonicity and
concavity provide identifying power. Point identification is achieved when the
structural function is flat or linear with respect to the explanatory variable
over a given interval. We compute the bounds using real data and show that our
bounds are informative.
| stat |
Finite Mixtures of ERGMs for Modeling Ensembles of Networks | Ensembles of networks arise in many scientific fields, but there are few
statistical tools for inferring their generative processes, particularly in the
presence of both dyadic dependence and cross-graph heterogeneity. To fill in
this gap, we propose characterizing network ensembles via finite mixtures of
exponential family random graph models, a framework for parametric statistical
modeling of graphs that has been successful in explicitly modeling the complex
stochastic processes that govern the structure of edges in a network. Our
proposed modeling framework can also be used for applications such as
model-based clustering of ensembles of networks and density estimation for
complex graph distributions. We develop a Metropolis-within-Gibbs algorithm to
conduct fully Bayesian inference and adapt a version of deviance information
criterion for missing data models to choose the number of latent heterogeneous
generative mechanisms. Simulation studies show that the proposed procedure can
recover the true number of latent heterogeneous generative processes and
corresponding parameters. We demonstrate the utility of the proposed approach
using an ensemble of political co-voting networks among U.S. Senators.
| stat |
Single Point Transductive Prediction | Standard methods in supervised learning separate training and prediction: the
model is fit independently of any test points it may encounter. However, can
knowledge of the next test point $\mathbf{x}_{\star}$ be exploited to improve
prediction accuracy? We address this question in the context of linear
prediction, showing how techniques from semi-parametric inference can be used
transductively to combat regularization bias. We first lower bound the
$\mathbf{x}_{\star}$ prediction error of ridge regression and the Lasso,
showing that they must incur significant bias in certain test directions. We
then provide non-asymptotic upper bounds on the $\mathbf{x}_{\star}$ prediction
error of two transductive prediction rules. We conclude by showing the efficacy
of our methods on both synthetic and real data, highlighting the improvements
single point transductive prediction can provide in settings with distribution
shift.
| stat |
Data Amplification: A Unified and Competitive Approach to Property
Estimation | Estimating properties of discrete distributions is a fundamental problem in
statistical learning. We design the first unified, linear-time, competitive,
property estimator that for a wide class of properties and for all underlying
distributions uses just $2n$ samples to achieve the performance attained by the
empirical estimator with $n\sqrt{\log n}$ samples. This provides off-the-shelf,
distribution-independent, "amplification" of the amount of data available
relative to common-practice estimators.
We illustrate the estimator's practical advantages by comparing it to
existing estimators for a wide variety of properties and distributions. In most
cases, its performance with $n$ samples is even as good as that of the
empirical estimator with $n\log n$ samples, and for essentially all properties,
its performance is comparable to that of the best existing estimator designed
specifically for that property.
| stat |
Bayesian Bootstrap Spike-and-Slab LASSO | The impracticality of posterior sampling has prevented the widespread
adoption of spike-and-slab priors in high-dimensional applications. To
alleviate the computational burden, optimization strategies have been proposed
that quickly find local posterior modes. Trading off uncertainty quantification
for computational speed, these strategies have enabled spike-and-slab
deployments at scales that would be previously unfeasible. We build on one
recent development in this strand of work: the Spike-and-Slab LASSO procedure
of Ro\v{c}kov\'{a} and George (2018). Instead of optimization, however, we
explore multiple avenues for posterior sampling, some traditional and some new.
Intrigued by the speed of Spike-and-Slab LASSO mode detection, we explore the
possibility of sampling from an approximate posterior by performing MAP
optimization on many independently perturbed datasets. To this end, we explore
Bayesian bootstrap ideas and introduce a new class of jittered Spike-and-Slab
LASSO priors with random shrinkage targets. These priors are a key constituent
of the Bayesian Bootstrap Spike-and-Slab LASSO (BB-SSL) method proposed here.
BB-SSL turns fast optimization into approximate posterior sampling. Beyond its
scalability, we show that BB-SSL has a strong theoretical support. Indeed, we
find that the induced pseudo-posteriors contract around the truth at a
near-optimal rate in sparse normal-means and in high-dimensional regression. We
compare our algorithm to the traditional Stochastic Search Variable Selection
(under Laplace priors) as well as many state-of-the-art methods for shrinkage
priors. We show, both in simulations and on real data, that our method fares
superbly in these comparisons, often providing substantial computational gains.
| stat |
An Overview of Statistical Data Analysis | The use of statistical software in academia and enterprises has been evolving
over the last years. More often than not, students, professors, workers, and
users, in general, have all had, at some point, exposure to statistical
software. Sometimes, difficulties are felt when dealing with such type of
software. Very few persons have theoretical knowledge to clearly understand
software configurations or settings, and sometimes even the presented results.
Very often, the users are required by academies or enterprises to present
reports, without the time to explore or understand the results or tasks
required to do an optimal preparation of data or software settings. In this
work, we present a statistical overview of some theoretical concepts, to
provide fast access to some concepts.
| stat |
Estimation and Selection Properties of the LAD Fused Lasso Signal
Approximator | The fused lasso is an important method for signal processing when the hidden
signals are sparse and blocky. It is often used in combination with the squared
loss function. However, the squared loss is not suitable for heavy tail error
distributions nor is robust against outliers which arise often in practice. The
least absolute deviations (LAD) loss provides a robust alternative to the
squared loss. In this paper, we study the asymptotic properties of the fused
lasso estimator with the LAD loss for signal approximation. We refer to this
estimator as the LAD fused lasso signal approximator, or LAD-FLSA. We
investigate the estimation consistency properties of the LAD-FLSA and provide
sufficient conditions under which the LAD-FLSA is sign consistent. We also
construct an unbiased estimator for the degrees of freedom of the LAD-FLSA for
any given tuning parameters. Both simulation studies and real data analysis are
conducted to illustrate the performance of the LAD-FLSA and the effect of the
unbiased estimator of the degrees of freedom.
| stat |
Learning Optimized Risk Scores | Risk scores are simple classification models that let users make quick risk
predictions by adding and subtracting a few small numbers. These models are
widely used in medicine and criminal justice, but are difficult to learn from
data because they need to be calibrated, sparse, use small integer
coefficients, and obey application-specific operational constraints. In this
paper, we present a new machine learning approach to learn risk scores. We
formulate the risk score problem as a mixed integer nonlinear program, and
present a cutting plane algorithm for non-convex settings to efficiently
recover its optimal solution. We improve our algorithm with specialized
techniques to generate feasible solutions, narrow the optimality gap, and
reduce data-related computation. Our approach can fit risk scores in a way that
scales linearly in the number of samples, provides a certificate of optimality,
and obeys real-world constraints without parameter tuning or post-processing.
We benchmark the performance benefits of this approach through an extensive set
of numerical experiments, comparing to risk scores built using heuristic
approaches. We also discuss its practical benefits through a real-world
application where we build a customized risk score for ICU seizure prediction
in collaboration with the Massachusetts General Hospital.
| stat |
Weather event severity prediction using buoy data and machine learning | In this paper, we predict severity of extreme weather events (tropical
storms, hurricanes, etc.) using buoy data time series variables such as wind
speed and air temperature. The prediction/forecasting method is based on
various forecasting and machine learning models. The following steps are used.
Data sources for the buoys and weather events are identified, aggregated and
merged. For missing data imputation, we use Kalman filters as well as splines
for multivariate time series. Then, statistical tests are run to ascertain
increasing trends in weather event severity. Next, we use machine learning to
predict/forecast event severity using buoy variables, and report good
accuracies for the models built.
| stat |
Forward-Backward Splitting for Optimal Transport based Problems | Optimal transport aims to estimate a transportation plan that minimizes a
displacement cost. This is realized by optimizing the scalar product between
the sought plan and the given cost, over the space of doubly stochastic
matrices. When the entropy regularization is added to the problem, the
transportation plan can be efficiently computed with the Sinkhorn algorithm.
Thanks to this breakthrough, optimal transport has been progressively extended
to machine learning and statistical inference by introducing additional
application-specific terms in the problem formulation. It is however
challenging to design efficient optimization algorithms for optimal transport
based extensions. To overcome this limitation, we devise a general
forward-backward splitting algorithm based on Bregman distances for solving a
wide range of optimization problems involving a differentiable function with
Lipschitz-continuous gradient and a doubly stochastic constraint. We illustrate
the efficiency of our approach in the context of continuous domain adaptation.
Experiments show that the proposed method leads to a significant improvement in
terms of speed and performance with respect to the state of the art for domain
adaptation on a continually rotating distribution coming from the standard two
moon dataset.
| stat |
Regression-Based Bayesian Estimation and Structure Learning for
Nonparanormal Graphical Models | A nonparanormal graphical model is a semiparametric generalization of a
Gaussian graphical model for continuous variables in which it is assumed that
the variables follow a Gaussian graphical model only after some unknown smooth
monotone transformations. We consider a Bayesian approach to inference in a
nonparanormal graphical model in which we put priors on the unknown
transformations through a random series based on B-splines. We use a regression
formulation to construct the likelihood through the Cholesky decomposition on
the underlying precision matrix of the transformed variables and put shrinkage
priors on the regression coefficients. We apply a plug-in variational Bayesian
algorithm for learning the sparse precision matrix and compare the performance
to a posterior Gibbs sampling scheme in a simulation study. We finally apply
the proposed methods to a real data set. KEYWORDS:
| stat |
Generalizing causal inferences from randomized trials: counterfactual
and graphical identification | When engagement with a randomized trial is driven by factors that affect the
outcome or when trial engagement directly affects the outcome independent of
treatment, the average treatment effect among trial participants is unlikely to
generalize to a target population. In this paper, we use counterfactual and
graphical causal models to examine under what conditions we can generalize
causal inferences from a randomized trial to the target population of
trial-eligible individuals. We offer an interpretation of generalizability
analyses using the notion of a hypothetical intervention to "scale-up" trial
engagement to the target population. We consider the interpretation of
generalizability analyses when trial engagement does or does not directly
affect the outcome, highlight connections with censoring in longitudinal
studies, and discuss identification of the distribution of counterfactual
outcomes via g-formula computation and inverse probability weighting. Last, we
show how the methods can be extended to address time-varying treatments,
non-adherence, and censoring.
| stat |
Nested Conformal Prediction Sets for Classification with Applications to
Probation Data | Risk assessments to help inform criminal justice decisions have been used in
the United States since the 1920s. Over the past several years, statistical
learning risk algorithms have been introduced amid much controversy about
fairness, transparency and accuracy. In this paper, we focus on accuracy for a
large department of probation and parole that is considering a major revision
of its current, statistical learning risk methods. Because the content of each
offender's supervision is substantially shaped by a forecast of subsequent
conduct, forecasts have real consequences. Here we consider the probability
that risk forecasts are correct. We augment standard statistical learning
estimates of forecasting uncertainty (i.e., confusion tables) with uncertainty
estimates from nested conformal prediction sets. In a demonstration of concept
using data from the department of probation and parole, we show that the
standard uncertainty measures and uncertainty measures from nested conformal
prediction sets can differ dramatically in concept and output. We also provide
a modification of nested conformal called the localized conformal method to
match confusion tables more closely when possible. A strong case can be made
favoring the nested and localized conformal approach. As best we can tell, our
formulation of such comparisons and consequent recommendations is novel.
| stat |
A Nonparametric Bayesian Design for Drug Combination Cancer Trials | We propose an adaptive design for early phase drug combination cancer trials
with the goal of estimating the maximum tolerated dose (MTD). A nonparametric
Bayesian model, using beta priors truncated to the set of partially ordered
dose combinations, is used to describe the probability of dose limiting
toxicity (DLT). Dose allocation between successive cohorts of patients is
estimated using a modified Continual Reassessment scheme. The updated
probabilities of DLT are calculated with a Gibbs sampler that employs a
weighting mechanism to calibrate the influence of data versus the prior. At the
end of the trial, we recommend one or more dose combinations as the MTD based
on our proposed algorithm. The design operating characteristics indicate that
our method is comparable with existing methods. As an illustration, we apply
our method to a phase I clinical trial of CB-839 and Gemcitabine.
| stat |
Optimal Cox Regression Subsampling Procedure with Rare Events | Massive sized survival datasets are becoming increasingly prevalent with the
development of the healthcare industry. Such datasets pose computational
challenges unprecedented in traditional survival analysis use-cases. A popular
way for coping with massive datasets is downsampling them to a more manageable
size, such that the computational resources can be afforded by the researcher.
Cox proportional hazards regression has remained one of the most popular
statistical models for the analysis of survival data to-date. This work
addresses the settings of right censored and possibly left truncated data with
rare events, such that the observed failure times constitute only a small
portion of the overall sample. We propose Cox regression subsampling-based
estimators that approximate their full-data partial-likelihood-based
counterparts, by assigning optimal sampling probabilities to censored
observations, and including all observed failures in the analysis. Asymptotic
properties of the proposed estimators are established under suitable regularity
conditions, and simulation studies are carried out to evaluate the finite
sample performance of the estimators. We further apply our procedure on
UK-biobank colorectal cancer genetic and environmental risk factors.
| stat |
Counterfactual diagnosis | Machine learning promises to revolutionize clinical decision making and
diagnosis. In medical diagnosis a doctor aims to explain a patient's symptoms
by determining the diseases \emph{causing} them. However, existing diagnostic
algorithms are purely associative, identifying diseases that are strongly
correlated with a patients symptoms and medical history. We show that this
inability to disentangle correlation from causation can result in sub-optimal
or dangerous diagnoses. To overcome this, we reformulate diagnosis as a
counterfactual inference task and derive new counterfactual diagnostic
algorithms. We show that this approach is closer to the diagnostic reasoning of
clinicians and significantly improves the accuracy and safety of the resulting
diagnoses. We compare our counterfactual algorithm to the standard Bayesian
diagnostic algorithm and a cohort of 44 doctors using a test set of clinical
vignettes. While the Bayesian algorithm achieves an accuracy comparable to the
average doctor, placing in the top 48% of doctors in our cohort, our
counterfactual algorithm places in the top 25% of doctors, achieving expert
clinical accuracy. This improvement is achieved simply by changing how we query
our model, without requiring any additional model improvements. Our results
show that counterfactual reasoning is a vital missing ingredient for applying
machine learning to medical diagnosis.
| stat |
Learning deep representations by mutual information estimation and
maximization | In this work, we perform unsupervised learning of representations by
maximizing mutual information between an input and the output of a deep neural
network encoder. Importantly, we show that structure matters: incorporating
knowledge about locality of the input to the objective can greatly influence a
representation's suitability for downstream tasks. We further control
characteristics of the representation by matching to a prior distribution
adversarially. Our method, which we call Deep InfoMax (DIM), outperforms a
number of popular unsupervised learning methods and competes with
fully-supervised learning on several classification tasks. DIM opens new
avenues for unsupervised learning of representations and is an important step
towards flexible formulations of representation-learning objectives for
specific end-goals.
| stat |
Treed distributed lag nonlinear models | In studies of maternal exposure to air pollution a children's health outcome
is regressed on exposures observed during pregnancy. The distributed lag
nonlinear model (DLNM) is a statistical method commonly implemented to estimate
an exposure-time-response function when it is postulated the exposure effect is
nonlinear. Previous implementations of the DLNM estimate an
exposure-time-response surface parameterized with a bivariate basis expansion.
However, basis functions such as splines assume smoothness across the entire
exposure-time-response surface, which may be unrealistic in settings where the
exposure is associated with the outcome only in a specific time window. We
propose a framework for estimating the DLNM based on Bayesian additive
regression trees. Our method operates using a set of regression trees that each
assume piecewise constant relationships across the exposure-time space. In a
simulation, we show that our model outperforms spline-based models when the
exposure-time surface is not smooth, while both methods perform similarly in
settings where the true surface is smooth. Importantly, the proposed approach
is lower variance and more precisely identifies critical windows during which
exposure is associated with a future health outcome. We apply our method to
estimate the association between maternal exposure to PM$_{2.5}$ and birth
weight in a Colorado USA birth cohort.
| stat |
Accelerating Langevin Sampling with Birth-death | A fundamental problem in Bayesian inference and statistical machine learning
is to efficiently sample from multimodal distributions. Due to metastability,
multimodal distributions are difficult to sample using standard Markov chain
Monte Carlo methods. We propose a new sampling algorithm based on a birth-death
mechanism to accelerate the mixing of Langevin diffusion. Our algorithm is
motivated by its mean field partial differential equation (PDE), which is a
Fokker-Planck equation supplemented by a nonlocal birth-death term. This PDE
can be viewed as a gradient flow of the Kullback-Leibler divergence with
respect to the Wasserstein-Fisher-Rao metric. We prove that under some
assumptions the asymptotic convergence rate of the nonlocal PDE is independent
of the potential barrier, in contrast to the exponential dependence in the case
of the Langevin diffusion. We illustrate the efficiency of the birth-death
accelerated Langevin method through several analytical examples and numerical
experiments.
| stat |
Rapid Detection of Hot-spots via Tensor Decomposition with applications
to Crime Rate Data | We propose an efficient statistical method (denoted as SSR-Tensor) to
robustly and quickly detect hot-spots that are sparse and temporal-consistent
in a spatial-temporal dataset through the tensor decomposition. Our main idea
is first to build an SSR model to decompose the tensor data into a Smooth
global trend mean, Sparse local hot-spots, and Residuals. Next, tensor
decomposition is utilized as follows: bases are introduced to describe
within-dimension correlation, and tensor products are used for
between-dimension interaction. Then, a combination of LASSO and fused LASSO is
used to estimate the model parameters, where an efficient recursive estimation
procedure is developed based on the large-scale convex optimization, where we
first transform the general LASSO optimization into regular LASSO optimization
and apply FISTA to solve it with the fastest convergence rate. Finally, a CUSUM
procedure is applied to detect when and where the hot-spot event occurs. We
compare the performance of the proposed method in a numerical simulation study
and a real-world case study, which contains a dataset including a collection of
three types of crime rates for U.S. mainland states during the year 1965-2014.
In both cases, the proposed SSR-Tensor is able to achieve the fast detection
and accurate localization of the hot-spots.
| stat |
Machine Learning by Unitary Tensor Network of Hierarchical Tree
Structure | The resemblance between the methods used in quantum-many body physics and in
machine learning has drawn considerable attention. In particular, tensor
networks (TNs) and deep learning architectures bear striking similarities to
the extent that TNs can be used for machine learning. Previous results used
one-dimensional TNs in image recognition, showing limited scalability and
flexibilities. In this work, we train two-dimensional hierarchical TNs to solve
image recognition problems, using a training algorithm derived from the
multi-scale entanglement renormalization ansatz. This approach introduces
mathematical connections among quantum many-body physics, quantum information
theory, and machine learning. While keeping the TN unitary in the training
phase, TN states are defined, which encode classes of images into quantum
many-body states. We study the quantum features of the TN states, including
quantum entanglement and fidelity. We find these quantities could be properties
that characterize the image classes, as well as the machine learning tasks.
| stat |
Deep Neural Generative Model of Functional MRI Images for Psychiatric
Disorder Diagnosis | Accurate diagnosis of psychiatric disorders plays a critical role in
improving the quality of life for patients and potentially supports the
development of new treatments. Many studies have been conducted on machine
learning techniques that seek brain imaging data for specific biomarkers of
disorders. These studies have encountered the following dilemma: A direct
classification overfits to a small number of high-dimensional samples but
unsupervised feature-extraction has the risk of extracting a signal of no
interest. In addition, such studies often provided only diagnoses for patients
without presenting the reasons for these diagnoses. This study proposed a deep
neural generative model of resting-state functional magnetic resonance imaging
(fMRI) data. The proposed model is conditioned by the assumption of the
subject's state and estimates the posterior probability of the subject's state
given the imaging data, using Bayes' rule. This study applied the proposed
model to diagnose schizophrenia and bipolar disorders. Diagnostic accuracy was
improved by a large margin over competitive approaches, namely classifications
of functional connectivity, discriminative/generative models of region-wise
signals, and those with unsupervised feature-extractors. The proposed model
visualizes brain regions largely related to the disorders, thus motivating
further biological investigation.
| stat |
Variance-based sensitivity analysis for time-dependent processes | The global sensitivity analysis of time-dependent processes requires
history-aware approaches. We develop for that purpose a variance-based method
that leverages the correlation structure of the problems under study and
employs surrogate models to accelerate the computations. The errors resulting
from fixing unimportant uncertain parameters to their nominal values are
analyzed through a priori estimates. We illustrate our approach on a harmonic
oscillator example and on a nonlinear dynamic cholera model.
| stat |
Bayesian inference of chaotic dynamics by merging data assimilation,
machine learning and expectation-maximization | The reconstruction from observations of high-dimensional chaotic dynamics
such as geophysical flows is hampered by (i) the partial and noisy observations
that can realistically be obtained, (ii) the need to learn from long time
series of data, and (iii) the unstable nature of the dynamics. To achieve such
inference from the observations over long time series, it has been suggested to
combine data assimilation and machine learning in several ways. We show how to
unify these approaches from a Bayesian perspective using
expectation-maximization and coordinate descents. In doing so, the model, the
state trajectory and model error statistics are estimated all together.
Implementations and approximations of these methods are discussed. Finally, we
numerically and successfully test the approach on two relevant low-order
chaotic models with distinct identifiability.
| stat |
Bayesian inference and non-linear extensions of the CIRCE method for
quantifying the uncertainty of closure relationships integrated into
thermal-hydraulic system codes | Uncertainty Quantification of closure relationships integrated into
thermal-hydraulic system codes is a critical prerequisite in applying the
Best-Estimate Plus Uncertainty (BEPU) methodology for nuclear safety and
licensing processes.The purpose of the CIRCE method is to estimate the
(log)-Gaussian probability distribution of a multiplicative factor applied to a
reference closure relationship in order to assess its uncertainty. Even though
this method has been implemented with success in numerous physical scenarios,
it can still suffer from substantial limitations such as the linearity
assumption and the difficulty of properly taking into account the inherent
statistical uncertainty. In the paper, we will extend the CIRCE method in two
aspects. On the one hand, we adopt the Bayesian setting putting prior
probability distributions on the parameters of the (log)-Gaussian distribution.
The posterior distribution of the parameters is then computed with respect to
an experimental database by means of Markov Chain Monte Carlo (MCMC)
algorithms. On the other hand, we tackle the more general setting where the
simulations do not move linearly against the multiplicative factor(s). MCMC
algorithms then become time-prohibitive when the thermal-hydraulic simulations
exceed a few minutes. This handicap is overcome by using Gaussian process (GP)
emulators which can yield both reliable and fast predictions of the
simulations. The GP-based MCMC algorithms will be applied to quantify the
uncertainty of two condensation closure relationships at a safety injection
with respect to a database of experimental tests. The thermal-hydraulic
simulations will be run with the CATHARE 2 computer code.
| stat |
Understanding Graph Neural Networks with Asymmetric Geometric Scattering
Transforms | The scattering transform is a multilayered wavelet-based deep learning
architecture that acts as a model of convolutional neural networks. Recently,
several works have introduced generalizations of the scattering transform for
non-Euclidean settings such as graphs. Our work builds upon these constructions
by introducing windowed and non-windowed graph scattering transforms based upon
a very general class of asymmetric wavelets. We show that these asymmetric
graph scattering transforms have many of the same theoretical guarantees as
their symmetric counterparts. This work helps bridge the gap between scattering
and other graph neural networks by introducing a large family of networks with
provable stability and invariance guarantees. This lays the groundwork for
future deep learning architectures for graph-structured data that have learned
filters and also provably have desirable theoretical properties.
| stat |
The effect of Hybrid Principal Components Analysis on the Signal
Compression Functional Regression: With EEG-fMRI Application | Objective: In some situations that exist both scalar and functional data,
called mixed and hybrid data, the hybrid PCA (HPCA) was introduced. Among the
regression models for the hybrid data, we can count covariate-adjusted HPCA,
the Semi-functional partial linear regression, function-on-function (FOF)
regression with signal compression, and functional additive regression, models.
In this article, we study the effects of HPCA decomposition of hybrid data on
the prediction accuracy of the FOF regression with signal compressions. Method:
We stated a two-step procedure for incorporating the HPCA in the functional
regressions. The first step is reconstructing the data based on the HPCAs and
the second step is merging data on the other dimensions and calculate the
point-wise average of the desired functional dimension. We also choose the
number of HPCA based on Mean Squared Perdition Error (MSPE). Result: In the two
simulations, we show that the regression models with the first HPCA have the
best accuracy prediction and model fit summaries among no HPCA and all HPCAs
with a training/testing approach. Finally, we applied our methodology to the
EEG-fMRI dataset. Conclusions: We conclude that our methodology improves the
prediction of the experiments with the EEG datasets. And we recommend that
instead of using the functional PCA on the desired dimension, reconstruct the
data with HPCA and average it on the other two dimensions for functional
regression models.
| stat |
Avoiding Resentment Via Monotonic Fairness | Classifiers that achieve demographic balance by explicitly using protected
attributes such as race or gender are often politically or culturally
controversial due to their lack of individual fairness, i.e. individuals with
similar qualifications will receive different outcomes. Individually and group
fair decision criteria can produce counter-intuitive results, e.g. that the
optimal constrained boundary may reject intuitively better candidates due to
demographic imbalance in similar candidates. Both approaches can be seen as
introducing individual resentment, where some individuals would have received a
better outcome if they either belonged to a different demographic class and had
the same qualifications, or if they remained in the same class but had
objectively worse qualifications (e.g. lower test scores). We show that both
forms of resentment can be avoided by using monotonically constrained machine
learning models to create individually fair, demographically balanced
classifiers.
| stat |
Gradient Boosts the Approximate Vanishing Ideal | In the last decade, the approximate vanishing ideal and its basis
construction algorithms have been extensively studied in computer algebra and
machine learning as a general model to reconstruct the algebraic variety on
which noisy data approximately lie. In particular, the basis construction
algorithms developed in machine learning are widely used in applications across
many fields because of their monomial-order-free property; however, they lose
many of the theoretical properties of computer-algebraic algorithms. In this
paper, we propose general methods that equip monomial-order-free algorithms
with several advantageous theoretical properties. Specifically, we exploit the
gradient to (i) sidestep the spurious vanishing problem in polynomial time to
remove symbolically trivial redundant bases, (ii) achieve consistent output
with respect to the translation and scaling of input, and (iii) remove
nontrivially redundant bases. The proposed methods work in a fully numerical
manner, whereas existing algorithms require the awkward monomial order or
exponentially costly (and mostly symbolic) computation to realize properties
(i) and (iii). To our knowledge, property (ii) has not been achieved by any
existing basis construction algorithm of the approximate vanishing ideal.
| stat |
Learning Bayesian Networks from Ordinal Data | Bayesian networks are a powerful framework for studying the dependency
structure of variables in a complex system. The problem of learning Bayesian
networks is tightly associated with the given data type. Ordinal data, such as
stages of cancer, rating scale survey questions, and letter grades for exams,
are ubiquitous in applied research. However, existing solutions are mainly for
continuous and nominal data. In this work, we propose an iterative
score-and-search method - called the Ordinal Structural EM (OSEM) algorithm -
for learning Bayesian networks from ordinal data. Unlike traditional approaches
designed for nominal data, we explicitly respect the ordering amongst the
categories. More precisely, we assume that the ordinal variables originate from
marginally discretizing a set of Gaussian variables, whose structural
dependence in the latent space follows a directed acyclic graph. Then, we adopt
the Structural EM algorithm and derive closed-form scoring functions for
efficient graph searching. Through simulation studies, we illustrate the
superior performance of the OSEM algorithm compared to the alternatives and
analyze various factors that may influence the learning accuracy. Finally, we
demonstrate the practicality of our method with a real-world application on
psychological survey data from 408 patients with co-morbid symptoms of
obsessive-compulsive disorder and depression.
| stat |
Bayesian Attention Modules | Attention modules, as simple and effective tools, have not only enabled deep
neural networks to achieve state-of-the-art results in many domains, but also
enhanced their interpretability. Most current models use deterministic
attention modules due to their simplicity and ease of optimization. Stochastic
counterparts, on the other hand, are less popular despite their potential
benefits. The main reason is that stochastic attention often introduces
optimization issues or requires significant model changes. In this paper, we
propose a scalable stochastic version of attention that is easy to implement
and optimize. We construct simplex-constrained attention distributions by
normalizing reparameterizable distributions, making the training process
differentiable. We learn their parameters in a Bayesian framework where a
data-dependent prior is introduced for regularization. We apply the proposed
stochastic attention modules to various attention-based models, with
applications to graph node classification, visual question answering, image
captioning, machine translation, and language understanding. Our experiments
show the proposed method brings consistent improvements over the corresponding
baselines.
| stat |
An Adaptive Weighted Deep Forest Classifier | A modification of the confidence screening mechanism based on adaptive
weighing of every training instance at each cascade level of the Deep Forest is
proposed. The idea underlying the modification is very simple and stems from
the confidence screening mechanism idea proposed by Pang et al. to simplify the
Deep Forest classifier by means of updating the training set at each level in
accordance with the classification accuracy of every training instance.
However, if the confidence screening mechanism just removes instances from
training and testing processes, then the proposed modification is more flexible
and assigns weights by taking into account the classification accuracy. The
modification is similar to the AdaBoost to some extent. Numerical experiments
illustrate good performance of the proposed modification in comparison with the
original Deep Forest proposed by Zhou and Feng.
| stat |
Covariance-engaged Classification of Sets via Linear Programming | Set classification aims to classify a set of observations as a whole, as
opposed to classifying individual observations separately. To formally
understand the unfamiliar concept of binary set classification, we first
investigate the optimal decision rule under the normal distribution, which
utilizes the empirical covariance of the set to be classified. We show that the
number of observations in the set plays a critical role in bounding the Bayes
risk. Under this framework, we further propose new methods of set
classification. For the case where only a few parameters of the model drive the
difference between two classes, we propose a computationally-efficient approach
to parameter estimation using linear programming, leading to the
Covariance-engaged LInear Programming Set (CLIPS) classifier. Its theoretical
properties are investigated for both independent case and various (short-range
and long-range dependent) time series structures among observations within each
set. The convergence rates of estimation errors and risk of the CLIPS
classifier are established to show that having multiple observations in a set
leads to faster convergence rates, compared to the standard classification
situation in which there is only one observation in the set. The applicable
domains in which the CLIPS performs better than competitors are highlighted in
a comprehensive simulation study. Finally, we illustrate the usefulness of the
proposed methods in classification of real image data in histopathology.
| stat |
Fast and Fair Simultaneous Confidence Bands for Functional Parameters | Quantifying uncertainty using confidence regions is a central goal of
statistical inference. Despite this, methodologies for confidence bands in
Functional Data Analysis are underdeveloped compared to estimation and
hypothesis testing. This work represents a major leap forward in this area by
presenting a new methodology for constructing simultaneous confidence bands for
functional parameter estimates. These bands possess a number of striking
qualities: (1) they have a nearly closed-form expression, (2) they give nearly
exact coverage, (3) they have a finite sample correction, (4) they do not
require an estimate of the full covariance of the parameter estimate, and (5)
they can be constructed adaptively according to a desired criteria. One option
for choosing bands we find especially interesting is the concept of fair bands
which allows us to do fair (or equitable) inference over subintervals and could
be especially useful in longitudinal studies over long time scales. Our bands
are constructed by integrating and extending tools from Random Field Theory, an
area that has yet to overlap with Functional Data Analysis.
| stat |
Are All Layers Created Equal? | Understanding deep neural networks has been a major research objective in
recent years with notable theoretical progress. A focal point of those studies
stems from the success of excessively large networks which defy the classical
wisdom of uniform convergence and learnability. We study empirically the
layer-wise functional structure of overparameterized deep models. We provide
evidence for the heterogeneous characteristic of layers. To do so, we introduce
the notion of robustness to post-training re-initialization and
re-randomization. We show that the layers can be categorized as either
``ambient'' or ``critical''. Resetting the ambient layers to their initial
values has no negative consequence, and in many cases they barely change
throughout training. On the contrary, resetting the critical layers completely
destroys the predictor and the performance drops to chanceh. Our study provides
further evidence that mere parameter counting or norm accounting is too coarse
in studying generalization of deep models, and flatness or robustness analysis
of the models needs to respect the network architectures.
| stat |
Statistical post-processing of wind speed forecasts using convolutional
neural networks | Current statistical post-processing methods for probabilistic weather
forecasting are not capable of using full spatial patterns from the numerical
weather prediction (NWP) model. In this paper we incorporate spatial wind speed
information by using convolutional neural networks (CNNs) and obtain
probabilistic wind speed forecasts in the Netherlands for 48 hours ahead, based
on KNMI's deterministic Harmonie-Arome NWP model. The probabilistic forecasts
from the CNNs are shown to have higher Brier skill scores for medium to higher
wind speeds, as well as a better continuous ranked probability score (CRPS) and
logarithmic score, than the forecasts from fully connected neural networks and
quantile regression forests. As a secondary result, we have compared the CNNs
using 3 different density estimation methods (quantized softmax (QS), kernel
mixture networks, and fitting a truncated normal distribution), and found the
probabilistic forecasts based on the QS method to be best.
| stat |
Statistical Exploration of Relationships Between Routine and Agnostic
Features Towards Interpretable Risk Characterization | As is typical in other fields of application of high throughput systems,
radiology is faced with the challenge of interpreting increasingly
sophisticated predictive models such as those derived from radiomics analyses.
Interpretation may be guided by the learning output from machine learning
models, which may however vary greatly with each technique. Whatever this
output model, it will raise some essential questions. How do we interpret the
prognostic model for clinical implementation? How can we identify potential
information structures within sets of radiomic features, in order to create
clinically interpretable models? And how can we recombine or exploit potential
relationships between features towards improved interpretability? A number of
statistical techniques are explored to assess (possibly nonlinear)
relationships between radiological features from different angles.
| stat |
Supervised Dimensionality Reduction for Big Data | To solve key biomedical problems, experimentalists now routinely measure
millions or billions of features (dimensions) per sample, with the hope that
data science techniques will be able to build accurate data-driven inferences.
Because sample sizes are typically orders of magnitude smaller than the
dimensionality of these data, valid inferences require finding a
low-dimensional representation that preserves the discriminating information
(e.g., whether the individual suffers from a particular disease). There is a
lack of interpretable supervised dimensionality reduction methods that scale to
millions of dimensions with strong statistical theoretical guarantees.We
introduce an approach, XOX, to extending principal components analysis by
incorporating class-conditional moment estimates into the low-dimensional
projection. The simplest ver-sion, "Linear Optimal Low-rank" projection (LOL),
incorporates the class-conditional means. We prove, and substantiate with both
synthetic and real data benchmarks, that LOL and its generalizations in the XOX
framework lead to improved data representations for subsequent classification,
while maintaining computational efficiency and scalability. Using multiple
brain imaging datasets consisting of >150 million features, and several
genomics datasets with>500,000 features, LOL outperforms other scalable linear
dimensionality reduction techniques in terms of accuracy, while only requiring
a few minutes on a standard desktop computer.
| stat |
A Comparison of Prior Elicitation Aggregation using the Classical Method
and SHELF | Subjective Bayesian prior distributions elicited from experts can be
aggregated together to form group priors. This paper compares aggregated priors
formed by Equal Weight Aggregation, the Classical Method, and the Sheffield
Elicitation Framework to each other and individual expert priors, using an
expert elicitation carried out for a clinical trial. Aggregation methods and
individual expert prior distributions are compared using proper scoring rules
to compare the informativeness and calibration of the distributions. The three
aggregation methods outperform the individual experts, and the Sheffield
Elicitation Framework performs best amongst them.
| stat |
Average Treatment Effects in the Presence of Interference | We propose a definition for the average indirect effect of a binary treatment
in the potential outcomes model for causal inference. Our definition is
analogous to the standard definition of the average direct effect, and can be
expressed without needing to compare outcomes across multiple randomized
experiments. We show that the proposed indirect effect satisfies a universal
decomposition theorem, whereby the sum of the average direct and indirect
effects always corresponds to the average effect of a policy intervention. We
also consider a number of parametric models for interference considered by
applied researchers, and find that our (non-parametrically defined) indirect
effect remains a natural estimand when re-expressed in the context of these
models.
| stat |
Passive Approach for the K-means Problem on Streaming Data | Currently the amount of data produced worldwide is increasing beyond measure,
thus a high volume of unsupervised data must be processed continuously. One of
the main unsupervised data analysis is clustering. In streaming data scenarios,
the data is composed by an increasing sequence of batches of samples where the
concept drift phenomenon may happen. In this paper, we formally define the
Streaming $K$-means(S$K$M) problem, which implies a restart of the error
function when a concept drift occurs. We propose a surrogate error function
that does not rely on concept drift detection. We proof that the surrogate is a
good approximation of the S$K$M error. Hence, we suggest an algorithm which
minimizes this alternative error each time a new batch arrives. We present some
initialization techniques for streaming data scenarios as well. Besides
providing theoretical results, experiments demonstrate an improvement of the
converged error for the non-trivial initialization methods.
| stat |
Homogeneity Tests of Covariance and Change-Points Identification for
High-Dimensional Functional Data | We consider inference problems for high-dimensional (HD) functional data with
a dense number (T) of repeated measurements taken for a large number of p
variables from a small number of n experimental units. The spatial and temporal
dependence, high dimensionality, and the dense number of repeated measurements
all make theoretical studies and computation challenging. This paper has two
aims; our first aim is to solve the theoretical and computational challenges in
detecting and identifying change points among covariance matrices from HD
functional data. The second aim is to provide computationally efficient and
tuning-free tools with a guaranteed stochastic error control. The change point
detection procedure is developed in the form of testing the homogeneity of
covariance matrices. The weak convergence of the stochastic process formed by
the test statistics is established under the "large p, large T and small n"
setting. Under a mild set of conditions, our change point identification
estimator is proven to be consistent for change points in any location of a
sequence. Its rate of convergence depends on the data dimension, sample size,
number of repeated measurements, and signal-to-noise ratio. We also show that
our proposed computation algorithms can significantly reduce the computation
time and are applicable to real-world data such as fMRI data with a large
number of HD repeated measurements. Simulation results demonstrate both finite
sample performance and computational effectiveness of our proposed procedures.
We observe that the empirical size of the test is well controlled at the
nominal level, and the locations of multiple change points can accurately be
identified. An application to fMRI data demonstrates that our proposed methods
can identify event boundaries in the preface of the movie Sherlock. Our
proposed procedures are implemented in an R package TechPhD.
| stat |
Random Intersection Chains | Interactions between several features sometimes play an important role in
prediction tasks. But taking all the interactions into consideration will lead
to an extremely heavy computational burden. For categorical features, the
situation is more complicated since the input will be extremely
high-dimensional and sparse if one-hot encoding is applied. Inspired by
association rule mining, we propose a method that selects interactions of
categorical features, called Random Intersection Chains. It uses random
intersections to detect frequent patterns, then selects the most meaningful
ones among them. At first a number of chains are generated, in which each node
is the intersection of the previous node and a random chosen observation. The
frequency of patterns in the tail nodes is estimated by maximum likelihood
estimation, then the patterns with largest estimated frequency are selected.
After that, their confidence is calculated by Bayes' theorem. The most
confident patterns are finally returned by Random Intersection Chains. We show
that if the number and length of chains are appropriately chosen, the patterns
in the tail nodes are indeed the most frequent ones in the data set. We analyze
the computation complexity of the proposed algorithm and prove the convergence
of the estimators. The results of a series of experiments verify the efficiency
and effectiveness of the algorithm.
| stat |
Bayesian Joint Modeling of Multiple Brain Functional Networks | Brain function is organized in coordinated modes of spatio-temporal activity
(functional networks) exhibiting an intrinsic baseline structure with
variations under different experimental conditions. Existing approaches for
uncovering such network structures typically do not explicitly model shared and
differential patterns across networks, thus potentially reducing the detection
power. We develop an integrative modeling approach for jointly modeling
multiple brain networks across experimental conditions. The proposed Bayesian
Joint Network Learning approach develops flexible priors on the edge
probabilities involving a common intrinsic baseline structure and differential
effects specific to individual networks. Conditional on these edge
probabilities, connection strengths are modeled under a Bayesian spike and slab
prior on the off-diagonal elements of the inverse covariance matrix. The model
is fit under a posterior computation scheme based on Markov chain Monte Carlo.
Numerical simulations illustrate that the proposed joint modeling approach has
increased power to detect true differential edges while providing adequate
control on false positives and achieving greater accuracy in the estimation of
edge strengths compared to existing methods. An application of the method to
fMRI Stroop task data provides unique insights into brain network alterations
between cognitive conditions which existing graphical modeling techniques
failed to reveal.
| stat |
Approaches Toward the Bayesian Estimation of the Stochastic Volatility
Model with Leverage | The sampling efficiency of MCMC methods in Bayesian inference for stochastic
volatility (SV) models is known to highly depend on the actual parameter
values, and the effectiveness of samplers based on different parameterizations
varies significantly. We derive novel algorithms for the centered and the
non-centered parameterizations of the practically highly relevant SV model with
leverage, where the return process and innovations of the volatility process
are allowed to correlate. Moreover, based on the idea of
ancillarity-sufficiency interweaving (ASIS), we combine the resulting samplers
in order to guarantee stable sampling efficiency irrespective of the baseline
parameterization.We carry out an extensive comparison to already existing
sampling methods for this model using simulated as well as real world data.
| stat |
Methods to Deal with Unknown Populational Minima during Parameter
Inference | There is a myriad of phenomena that are better modelled with semi-infinite
distribution families, many of which are studied in survival analysis. When
performing inference, lack of knowledge of the populational minimum becomes a
problem, which can be dealt with by making a good guess thereof, or by
handcrafting a grid of initial parameters that will be useful for that
particular problem. These solutions are fine when analyzing a single set of
samples, but it becomes unfeasible when there are multiple datasets and a
case-by-case analysis would be too time consuming. In this paper we propose
methods to deal with the populational minimum in algorithmic, efficient and/or
simple ways. Six methods are presented and analyzed, two of which have full
theoretical support, but lack simplicity. The other four are simple and have
some theoretical grounds in non-parametric results such as the law of iterated
logarithm, and they exhibited very good results when it comes to maximizing
likelihood and being able to recycle the grid of initial parameters among the
datasets. With our results, we hope to ease the inference process for
practitioners, and expect that these methods will eventually be included in
software packages themselves.
| stat |
Seamless multi-model postprocessing for air temperature forecasts in
complex topography | Statistical postprocessing is routinely applied to correct systematic errors
of numerical weather prediction models (NWP) and to automatically produce
calibrated local forecasts for end-users. Postprocessing is particularly
relevant in complex terrain, where even state-of-the-art high-resolution NWP
systems cannot resolve many of the small-scale processes shaping local weather
conditions. In addition, statistical postprocessing can also be used to combine
forecasts from multiple NWP systems. Here we assess an ensemble model output
statistics (EMOS) approach to produce seamless temperature forecasts based on a
combination of short-term ensemble forecasts from a convection-permitting
limited-area ensemble and a medium-range global ensemble forecasting model. We
quantify the benefit of this approach compared to only processing the
high-resolution NWP. We calibrate and combine 2-m air temperature predictions
for a large set of Swiss weather stations at the hourly time-scale. The
multi-model EMOS approach ('Mixed EMOS') is able to improve forecasts by 30\%
with respect to direct model output from the high-resolution NWP. A detailed
evaluation of Mixed EMOS reveals that it outperforms either single-model EMOS
version by 8-12\%. Valley location profit particularly from the model
combination. All forecast variants perform worst in winter (DJF), however
calibration and model combination improves forecast quality substantially.
| stat |
Mixture of hidden Markov models for accelerometer data | Motivated by the analysis of accelerometer data, we introduce a specific
finite mixture of hidden Markov models with particular characteristics that
adapt well to the specific nature of this type of data. Our model allows for
the computation of statistics that characterize the physical activity of a
subject (\emph{e.g.}, the mean time spent at different activity levels and the
probability of the transition between two activity levels) without specifying
the activity levels in advance but by estimating them from the data. In
addition, this approach allows the heterogeneity of the population to be taken
into account and subpopulations with homogeneous physical activity behavior to
be defined. We prove that, under mild assumptions, this model implies that the
probability of misclassifying a subject decreases at an exponential decay with
the length of its measurement sequence. Model identifiability is also
investigated. We also report a comprehensive suite of numerical simulations to
support our theoretical findings. Method is motivated by and applied to the PAT
study.
| stat |
Simultaneous test for Means: An Unblind Way to the F-test in One-way
Analysis of Variance | After rejecting the null hypothesis in the analysis of variance, the next
step is to make the pairwise comparisons to find out differences in means. The
purpose of this paper is threefold. The foremost aim is to suggest expression
for calculating decision limit that enables us to collect the test and pairwise
comparisons in one step. This expression is proposed as the ratio of between
square for each treatment and within sum of squares for all treatments. The
second aim is to obtain the exact sampling distribution of the proposed ratio
under the null hypothesis. The exact sampling distribution is derived as the
beta distribution of the second type. The third aim is to use beta distribution
and adjusted p values to create decision limit. Therefore, reject the null
hypothesis of equal means if any adjusted point falls outsides the decision
limit. Simulation study is conducted to compute Type one error. The results
show that the proposed method controls the type one error near nominal values
using Benjamini-Hochberg adjusted p-values. Two applications are given to show
the benefits of the proposed method.
| stat |
Unit-root test within a threshold ARMA framework | We propose a new unit-root test based on Lagrange Multipliers, where we
extend the null hypothesis to an integrated moving-average process (IMA(1,1))
and the alternative to a first-order threshold autoregressive moving-average
process (TARMA(1,1)). This new theoretical framework provides tests with good
size without pre-modelling steps. Moreover, leveraging on the versatile
capability of the TARMA(1,1), our test has power against a wide range of linear
and nonlinear alternatives. We prove the consistency and asymptotic similarity
of the test. The proof of tightness of the test is of independent and general
theoretical interest. Moreover, we propose a wild bootstrap version of the
statistic. Our proposals outperform most existing tests in many contexts. We
support the view that rejection does not necessarily imply nonlinearity so that
unit-root tests should not be used uncritically to select a model. Finally, we
present an application to real exchange rates.
| stat |
Scalable Bayesian modeling for smoothing disease risks in large spatial
data sets | Several methods have been proposed in the spatial statistics literature for
the analysis of big data sets in continuous domains. However, new methods for
analyzing high-dimensional areal data are still scarce. Here, we propose a
scalable Bayesian modeling approach for smoothing mortality (or incidence)
risks in high-dimensional data, that is, when the number of small areas is very
large. The method is implemented in the R add-on package bigDM. Model fitting
and inference is based on the idea of "divide and conquer" and use integrated
nested Laplace approximations and numerical integration. We analyze the
proposal's empirical performance in a comprehensive simulation study that
consider two model-free settings. Finally, the methodology is applied to
analyze male colorectal cancer mortality in Spanish municipalities showing its
benefits with regard to the standard approach in terms of goodness of fit and
computational time.
| stat |
Dimensionality reduction to maximize prediction generalization
capability | This work develops an analytically solvable unsupervised learning scheme that
extracts the most informative components for predicting future inputs, termed
predictive principal component analysis (PredPCA). Our scheme can effectively
remove unpredictable observation noise and globally minimize the test
prediction error. Mathematical analyses demonstrate that, with sufficiently
high-dimensional observations that are generated by a linear or nonlinear
system, PredPCA can identify the optimal hidden state representation, true
system parameters, and true hidden state dimensionality, with a global
convergence guarantee. We demonstrate the performance of PredPCA by using
sequential visual inputs comprising hand-digits, rotating 3D objects, and
natural scenes. It reliably and accurately estimates distinct hidden states and
predicts future outcomes of previously unseen test input data, even in the
presence of considerable observation noise. The simple model structure and low
computational cost of PredPCA make it highly desirable as a learning scheme for
biological neural networks and neuromorphic chips.
| stat |
Permutation-Based Causal Structure Learning with Unknown Intervention
Targets | We consider the problem of estimating causal DAG models from a mix of
observational and interventional data, when the intervention targets are
partially or completely unknown. This problem is highly relevant for example in
genomics, since gene knockout technologies are known to have off-target
effects. We characterize the interventional Markov equivalence class of DAGs
that can be identified from interventional data with unknown intervention
targets. In addition, we propose a provably consistent algorithm for learning
the interventional Markov equivalence class from such data. The proposed
algorithm greedily searches over the space of permutations to minimize a novel
score function. The algorithm is nonparametric, which is particularly important
for applications to genomics, where the relationships between variables are
often non-linear and the distribution non-Gaussian. We demonstrate the
performance of our algorithm on synthetic and biological datasets. Links to an
implementation of our algorithm and to a reproducible code base for our
experiments can be found at https://uhlerlab.github.io/causaldag/utigsp.
| stat |
A Hierarchical Spike-and-Slab Model for Pan-Cancer Survival Using
Pan-Omic Data | Pan-omics, pan-cancer analysis has advanced our understanding of the
molecular heterogeneity of cancer, expanding what was known from single-cancer
or single-omics studies. However, pan-cancer, pan-omics analyses have been
limited in their ability to use information from multiple sources of data
(e.g., omics platforms) and multiple sample sets (e.g., cancer types) to
predict important clinical outcomes, like overall survival. We address the
issue of prediction across multiple high-dimensional sources of data and
multiple sample sets by using exploratory results from BIDIFAC+, a method for
integrative dimension reduction of bidimensionally-linked matrices, in a
predictive model. We apply a Bayesian hierarchical model that performs variable
selection using spike-and-slab priors which are modified to allow for the
borrowing of information across clustered data. This method is used to predict
overall patient survival from the Cancer Genome Atlas (TCGA) using data from 29
cancer types and 4 omics sources. Our model selected patterns of variation
identified by BIDIFAC+ that differentiate clinical tumor subtypes with markedly
different survival outcomes. We also use simulations to evaluate the
performance of the modified spike-and-slab prior in terms of its variable
selection accuracy and prediction accuracy under different underlying
data-generating frameworks. Software and code used for our analysis can be
found at https://github.com/sarahsamorodnitsky/HierarchicalSS_PanCanPanOmics/ .
| stat |
Domain Adaptation on Graphs by Learning Graph Topologies: Theoretical
Analysis and an Algorithm | Traditional machine learning algorithms assume that the training and test
data have the same distribution, while this assumption does not necessarily
hold in real applications. Domain adaptation methods take into account the
deviations in the data distribution. In this work, we study the problem of
domain adaptation on graphs. We consider a source graph and a target graph
constructed with samples drawn from data manifolds. We study the problem of
estimating the unknown class labels on the target graph using the label
information on the source graph and the similarity between the two graphs. We
particularly focus on a setting where the target label function is learnt such
that its spectrum is similar to that of the source label function. We first
propose a theoretical analysis of domain adaptation on graphs and present
performance bounds that characterize the target classification error in terms
of the properties of the graphs and the data manifolds. We show that the
classification performance improves as the topologies of the graphs get more
balanced, i.e., as the numbers of neighbors of different graph nodes become
more proportionate, and weak edges with small weights are avoided. Our results
also suggest that graph edges between too distant data samples should be
avoided for good generalization performance. We then propose a graph domain
adaptation algorithm inspired by our theoretical findings, which estimates the
label functions while learning the source and target graph topologies at the
same time. The joint graph learning and label estimation problem is formulated
through an objective function relying on our performance bounds, which is
minimized with an alternating optimization scheme. Experiments on synthetic and
real data sets suggest that the proposed method outperforms baseline
approaches.
| stat |
Online Learning Demands in Max-min Fairness | We describe mechanisms for the allocation of a scarce resource among multiple
users in a way that is efficient, fair, and strategy-proof, but when users do
not know their resource requirements. The mechanism is repeated for multiple
rounds and a user's requirements can change on each round. At the end of each
round, users provide feedback about the allocation they received, enabling the
mechanism to learn user preferences over time. Such situations are common in
the shared usage of a compute cluster among many users in an organisation,
where all teams may not precisely know the amount of resources needed to
execute their jobs. By understating their requirements, users will receive less
than they need and consequently not achieve their goals. By overstating them,
they may siphon away precious resources that could be useful to others in the
organisation. We formalise this task of online learning in fair division via
notions of efficiency, fairness, and strategy-proofness applicable to this
setting, and study this problem under three types of feedback: when the users'
observations are deterministic, when they are stochastic and follow a
parametric model, and when they are stochastic and nonparametric. We derive
mechanisms inspired by the classical max-min fairness procedure that achieve
these requisites, and quantify the extent to which they are achieved via
asymptotic rates. We corroborate these insights with an experimental evaluation
on synthetic problems and a web-serving task.
| stat |
Diversity of Artists in Major U.S. Museums | The U.S. art museum sector is grappling with diversity. While previous work
has investigated the demographic diversity of museum staffs and visitors, the
diversity of artists in their collections has remained unreported. We conduct
the first large-scale study of artist diversity in museums. By scraping the
public online catalogs of 18 major U.S. museums, deploying a sample of 10,000
artist records comprising over 9,000 unique artists to crowdsourcing, and
analyzing 45,000 responses, we infer artist genders, ethnicities, geographic
origins, and birth decades. Our results are threefold. First, we provide
estimates of gender and ethnic diversity at each museum, and overall, we find
that 85% of artists are white and 87% are men. Second, we identify museums that
are outliers, having significantly higher or lower representation of certain
demographic groups than the rest of the pool. Third, we find that the
relationship between museum collection mission and artist diversity is weak,
suggesting that a museum wishing to increase diversity might do so without
changing its emphases on specific time periods and regions. Our methodology can
be used to broadly and efficiently assess diversity in other fields.
| stat |
Estimating Individualized Decision Rules with Tail Controls | With the emergence of precision medicine, estimating optimal individualized
decision rules (IDRs) has attracted tremendous attention in many scientific
areas. Most existing literature has focused on finding optimal IDRs that can
maximize the expected outcome for each individual. Motivated by complex
individualized decision making procedures and popular conditional value at risk
(CVaR) measures, we propose a new robust criterion to estimate optimal IDRs in
order to control the average lower tail of the subjects' outcomes. In addition
to improving the individualized expected outcome, our proposed criterion takes
risks into consideration, and thus the resulting IDRs can prevent adverse
events. The optimal IDR under our criterion can be interpreted as the decision
rule that maximizes the ``worst-case" scenario of the individualized outcome
when the underlying distribution is perturbed within a constrained set. An
efficient non-convex optimization algorithm is proposed with convergence
guarantees. We investigate theoretical properties for our estimated optimal
IDRs under the proposed criterion such as consistency and finite sample error
bounds. Simulation studies and a real data application are used to further
demonstrate the robust performance of our method.
| stat |
Quantifying the uncertainty of variance partitioning estimates of
ecological datasets | An important objective of experimental biology is the quantification of the
relationship between predictor and response variables, a statistical analysis
often termed variance partitioning (VP). In this paper, a series of simulations
is presented, aiming to generate quantitative estimates of the expected
statistical uncertainty of VP analyses. We demonstrate scenarios with
considerable uncertainty in VP estimates, such that it significantly reduces
the statistical reliability of the obtained results. Especially when a
predictor variable of a dataset shows a low between-group variance, VP
estimates may show a high margin of error. This becomes particularly important
when the respective predictor variable only explains a small fraction of the
overall variance, or the number of replicates is particularly small. Moreover,
it is demonstrated that the expected error of VP estimates of a dataset can be
approximated by bootstrap resampling, giving researchers a tool for the
quantification of the uncertainty associated with an arbitrary VP analysis. The
applicability of this method is demonstrated by a re-analysis of the Oribatid
mite dataset introduced by Borcard and Legendre in 1994 and the Barro Colorado
Island tree count dataset by Condit and colleagues. We believe that this study
may encourage biologists to approach routine statistical analyses such as VP
more critically, and report the error associated with them more frequently.
| stat |
Sampling Conditionally on a Rare Event via Generalized Splitting | We propose and analyze a generalized splitting method to sample approximately
from a distribution conditional on the occurrence of a rare event. This has
important applications in a variety of contexts in operations research,
engineering, and computational statistics. The method uses independent trials
starting from a single particle. We exploit this independence to obtain
asymptotic and non-asymptotic bounds on the total variation error of the
sampler. Our main finding is that the approximation error depends crucially on
the relative variability of the number of points produced by the splitting
algorithm in one run, and that this relative variability can be readily
estimated via simulation. We illustrate the relevance of the proposed method on
an application in which one needs to sample (approximately) from an intractable
posterior density in Bayesian inference.
| stat |
Conditional Inference: Towards a Hierarchy of Statistical Evidence | Statistical uncertainty has many sources. P-values and confidence intervals
usually quantify the overall uncertainty, which may include variation due to
sampling and uncertainty due to measurement error, among others. Practitioners
might be interested in quantifying only one source of uncertainty. For example,
one might be interested in the uncertainty of a regression coefficient of a
fixed set of subjects, which corresponds to quantifying the uncertainty due to
measurement error and ignoring the variation induced by sampling. In causal
inference it is common to infer treatment effects for a certain set of
subjects, only accounting for uncertainty due to random treatment assignment.
Motivated by these examples, we consider conditional estimation and conditional
inference for parameters in parametric and semi-parametric models, where we
condition on observed characteristics of a population. We derive a theory of
conditional inference, including methods to obtain conditionally valid p-values
and confidence intervals. Conditional p- values can be used to construct a
hierarchy of statistical evidence that may help clarify the generalizability of
a statistical finding. We show that a naive method allows to gauge the
generalizability of a finding, with rigorous control of the family-wise error
rate. In addition, the proposed approach allows to conduct transfer learning of
conditional parameters, with rigorous conditional guarantees. The performance
of the proposed approach is evaluated on simulated and real-world data.
| stat |
Bayesian Low Rank Tensor Ring Model for Image Completion | Low rank tensor ring model is powerful for image completion which recovers
missing entries in data acquisition and transformation. The recently proposed
tensor ring (TR) based completion algorithms generally solve the low rank
optimization problem by alternating least squares method with predefined ranks,
which may easily lead to overfitting when the unknown ranks are set too large
and only a few measurements are available. In this paper, we present a Bayesian
low rank tensor ring model for image completion by automatically learning the
low rank structure of data. A multiplicative interaction model is developed for
the low-rank tensor ring decomposition, where core factors are enforced to be
sparse by assuming their entries obey Student-T distribution. Compared with
most of the existing methods, the proposed one is free of parameter-tuning, and
the TR ranks can be obtained by Bayesian inference. Numerical Experiments,
including synthetic data, color images with different sizes and YaleFace
dataset B with respect to one pose, show that the proposed approach outperforms
state-of-the-art ones, especially in terms of recovery accuracy.
| stat |
Efficient structure learning with automatic sparsity selection for
causal graph processes | We propose a novel algorithm for efficiently computing a sparse directed
adjacency matrix from a group of time series following a causal graph process.
Our solution is scalable for both dense and sparse graphs and automatically
selects the LASSO coefficient to obtain an appropriate number of edges in the
adjacency matrix. Current state-of-the-art approaches rely on
sparse-matrix-computation libraries to scale, and either avoid automatic
selection of the LASSO penalty coefficient or rely on the prediction mean
squared error, which is not directly related to the correct number of edges.
Instead, we propose a cyclical coordinate descent algorithm that employs two
new non-parametric error metrics to automatically select the LASSO coefficient.
We demonstrate state-of-the-art performance of our algorithm on simulated
stochastic block models and a real dataset of stocks from the S\&P$500$.
| stat |
Autoregressive Wild Bootstrap Inference for Nonparametric Trends | In this paper we propose an autoregressive wild bootstrap method to construct
confidence bands around a smooth deterministic trend. The bootstrap method is
easy to implement and does not require any adjustments in the presence of
missing data, which makes it particularly suitable for climatological
applications. We establish the asymptotic validity of the bootstrap method for
both pointwise and simultaneous confidence bands under general conditions,
allowing for general patterns of missing data, serial dependence and
heteroskedasticity. The finite sample properties of the method are studied in a
simulation study. We use the method to study the evolution of trends in daily
measurements of atmospheric ethane obtained from a weather station in the Swiss
Alps, where the method can easily deal with the many missing observations due
to adverse weather conditions.
| stat |
Feature Interaction Interpretability: A Case for Explaining
Ad-Recommendation Systems via Neural Interaction Detection | Recommendation is a prevalent application of machine learning that affects
many users; therefore, it is important for recommender models to be accurate
and interpretable. In this work, we propose a method to both interpret and
augment the predictions of black-box recommender systems. In particular, we
propose to interpret feature interactions from a source recommender model and
explicitly encode these interactions in a target recommender model, where both
source and target models are black-boxes. By not assuming the structure of the
recommender system, our approach can be used in general settings. In our
experiments, we focus on a prominent use of machine learning recommendation:
ad-click prediction. We found that our interaction interpretations are both
informative and predictive, e.g., significantly outperforming existing
recommender models. What's more, the same approach to interpret interactions
can provide new insights into domains even beyond recommendation, such as text
and image classification.
| stat |
Robust Estimation for Linear Panel Data Models | In different fields of applications including, but not limited to,
behavioral, environmental, medical sciences and econometrics, the use of panel
data regression models has become increasingly popular as a general framework
for making meaningful statistical inferences. However, when the ordinary least
squares (OLS) method is used to estimate the model parameters, presence of
outliers may significantly alter the adequacy of such models by producing
biased and inefficient estimates. In this work we propose a new, weighted
likelihood based robust estimation procedure for linear panel data models with
fixed and random effects. The finite sample performances of the proposed
estimators have been illustrated through an extensive simulation study as well
as with an application to blood pressure data set. Our thorough study
demonstrates that the proposed estimators show significantly better
performances over the traditional methods in the presence of outliers and
produce competitive results to the OLS based estimates when no outliers are
present in the data set.
| stat |
Bayesian Optimization Allowing for Common Random Numbers | Bayesian optimization is a powerful tool for expensive stochastic black-box
optimization problems such as simulation-based optimization or machine learning
hyperparameter tuning. Many stochastic objective functions implicitly require a
random number seed as input. By explicitly reusing a seed a user can exploit
common random numbers, comparing two or more inputs under the same randomly
generated scenario, such as a common customer stream in a job shop problem, or
the same random partition of training data into training and validation set for
a machine learning algorithm. With the aim of finding an input with the best
average performance over infinitely many seeds, we propose a novel Gaussian
process model that jointly models both the output for each seed and the
average. We then introduce the Knowledge gradient for Common Random Numbers
that iteratively determines a combination of input and random seed to evaluate
the objective and automatically trades off reusing old seeds and querying new
seeds, thus overcoming the need to evaluate inputs in batches or measuring
differences of pairs as suggested in previous methods. We investigate the
Knowledge Gradient for Common Random Numbers both theoretically and
empirically, finding it achieves significant performance improvements with only
moderate added computational cost.
| stat |
Descriptive and Predictive Analysis of Euroleague Basketball Games and
the Wisdom of Basketball Crowds | In this study we focus on the prediction of basketball games in the
Euroleague competition using machine learning modelling. The prediction is a
binary classification problem, predicting whether a match finishes 1 (home win)
or 2 (away win). Data is collected from the Euroleague's official website for
the seasons 2016-2017, 2017-2018 and 2018-2019, i.e. in the new format era.
Features are extracted from matches' data and off-the-shelf supervised machine
learning techniques are applied. We calibrate and validate our models. We find
that simple machine learning models give accuracy not greater than 67% on the
test set, worse than some sophisticated benchmark models. Additionally, the
importance of this study lies in the "wisdom of the basketball crowd" and we
demonstrate how the predicting power of a collective group of basketball
enthusiasts can outperform machine learning models discussed in this study. We
argue why the accuracy level of this group of "experts" should be set as the
benchmark for future studies in the prediction of (European) basketball games
using machine learning.
| stat |
Error-guided likelihood-free MCMC | This work presents a novel posterior inference method for models with
intractable evidence and likelihood functions. Error-guided likelihood-free
MCMC, or EG-LF-MCMC in short, has been developed for scientific applications,
where a researcher is interested in obtaining approximate posterior densities
over model parameters, while avoiding the need for expensive training of
component estimators on full observational data or the tedious design of
expressive summary statistics, as in related approaches. Our technique is based
on two phases. In the first phase, we draw samples from the prior, simulate
respective observations and record their errors $\epsilon$ in relation to the
true observation. We train a classifier to distinguish between corresponding
and non-corresponding $(\epsilon, \boldsymbol{\theta})$-tuples. In the second
stage the said classifier is conditioned on the smallest recorded $\epsilon$
value from the training set and employed for the calculation of transition
probabilities in a Markov Chain Monte Carlo sampling procedure. By conditioning
the MCMC on specific $\epsilon$ values, our method may also be used in an
amortized fashion to infer posterior densities for observations, which are
located a given distance away from the observed data. We evaluate the proposed
method on benchmark problems with semantically and structurally different data
and compare its performance against the state of the art approximate Bayesian
computation (ABC).
| stat |
On the normalized power prior | The power prior is a popular tool for constructing informative prior
distributions based on historical data. The method consists of raising the
likelihood to a discounting factor in order to control the amount of
information borrowed from the historical data. It is customary to perform a
sensitivity analysis reporting results for a range of values of the discounting
factor. However, one often wishes to assign it a prior distribution and
estimate it jointly with the parameters, which in turn necessitates the
computation of a normalising constant. In this paper we are concerned with how
to recycle computations from a sensitivity analysis in order to approximately
sample from joint posterior of the parameters and the discounting factor. We
first show a few important properties of the normalising constant and then use
these results to motivate a bisection-type algorithm for computing it on a
fixed budget of evaluations. We give a large array of illustrations and discuss
cases where the normalising constant is known in closed-form and where it is
not. We show that the proposed method produces approximate posteriors that are
very close to the exact distributions when those are available and also
produces posteriors that cover the data-generating parameters with higher
probability in the intractable case. Our results show that proper inclusion the
normalising constant is crucial to the correct quantification of uncertainty
and that the proposed method is an accurate and easy to implement technique to
include this normalisation, being applicable to a large class of models.
Key-words: Doubly-intractable; elicitation; historical data; normalisation;
power prior; sensitivity analysis.
| stat |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.