title
stringlengths 6
244
| abstract
stringlengths 19
5.09k
| label
stringclasses 10
values |
|---|---|---|
A Linear Bandit for Seasonal Environments | Contextual bandit algorithms are extremely popular and widely used in
recommendation systems to provide online personalised recommendations. A
recurrent assumption is the stationarity of the reward function, which is
rather unrealistic in most of the real-world applications. In the music
recommendation scenario for instance, people's music taste can abruptly change
during certain events, such as Halloween or Christmas, and revert to the
previous music taste soon after.
We would therefore need an algorithm which can promptly react to these
changes. Moreover, we would like to leverage already observed rewards collected
during different stationary periods which can potentially reoccur, without the
need of restarting the learning process from scratch. A growing literature has
addressed the problem of reward's non-stationarity, providing algorithms that
could quickly adapt to the changing environment. However, up to our knowledge,
there is no algorithm which deals with seasonal changes of the reward function.
Here we present a contextual bandit algorithm which detects and adapts to
abrupt changes of the reward function and leverages previous estimations
whenever the environment falls back to a previously observed state. We show
that the proposed method can outperform state-of-the-art algorithms for
non-stationary environments. We ran our experiment on both synthetic and real
datasets.
| stat |
Computer Model Calibration with Time Series Data using Deep Learning and
Quantile Regression | Computer models play a key role in many scientific and engineering problems.
One major source of uncertainty in computer model experiment is input parameter
uncertainty. Computer model calibration is a formal statistical procedure to
infer input parameters by combining information from model runs and
observational data. The existing standard calibration framework suffers from
inferential issues when the model output and observational data are
high-dimensional dependent data such as large time series due to the difficulty
in building an emulator and the non-identifiability between effects from input
parameters and data-model discrepancy. To overcome these challenges we propose
a new calibration framework based on a deep neural network (DNN) with
long-short term memory layers that directly emulates the inverse relationship
between the model output and input parameters. Adopting the 'learning with
noise' idea we train our DNN model to filter out the effects from data model
discrepancy on input parameter inference. We also formulate a new way to
construct interval predictions for DNN using quantile regression to quantify
the uncertainty in input parameter estimates. Through a simulation study and
real data application with WRF-hydro model we show that our approach can yield
accurate point estimates and well calibrated interval estimates for input
parameters.
| stat |
Integer-Valued Functional Data Analysis for Measles Forecasting | Measles presents a unique and imminent challenge for epidemiologists and
public health officials: the disease is highly contagious, yet vaccination
rates are declining precipitously in many localities. Consequently, the risk of
a measles outbreak continues to rise. To improve preparedness, we study
historical measles data both pre- and post-vaccine, and design new methodology
to forecast measles counts with uncertainty quantification. We propose to model
the disease counts as an integer-valued functional time series: measles counts
are a function of time-of-year and time-ordered by year. The counts are modeled
using a negative-binomial distribution conditional on a real-valued latent
process, which accounts for the overdispersion observed in the data. The latent
process is decomposed using an unknown basis expansion, which is learned from
the data, with dynamic basis coefficients. The resulting framework provides
enhanced capability to model complex seasonality, which varies dynamically from
year-to-year, and offers improved multi-month ahead point forecasts and
substantially tighter forecast intervals (with correct coverage) compared to
existing forecasting models. Importantly, the fully Bayesian approach provides
well-calibrated and precise uncertainty quantification for epi-relevent
features, such as the future value and time of the peak measles count in a
given year. An R package is available online.
| stat |
A Locally Adaptive Bayesian Cubature Method | Bayesian cubature (BC) is a popular inferential perspective on the cubature
of expensive integrands, wherein the integrand is emulated using a stochastic
process model. Several approaches have been put forward to encode sequential
adaptation (i.e. dependence on previous integrand evaluations) into this
framework. However, these proposals have been limited to either estimating the
parameters of a stationary covariance model or focusing computational resources
on regions where large values are taken by the integrand. In contrast, many
classical adaptive cubature methods focus computational resources on spatial
regions in which local error estimates are largest. The contributions of this
work are three-fold: First, we present a theoretical result that suggests there
does not exist a direct Bayesian analogue of the classical adaptive trapezoidal
method. Then we put forward a novel BC method that has empirically similar
behaviour to the adaptive trapezoidal method. Finally we present evidence that
the novel method provides improved cubature performance, relative to standard
BC, in a detailed empirical assessment.
| stat |
A Central Limit Theorem for Differentially Private Query Answering | Perhaps the single most important use case for differential privacy is to
privately answer numerical queries, which is usually achieved by adding noise
to the answer vector. The central question, therefore, is to understand which
noise distribution optimizes the privacy-accuracy trade-off, especially when
the dimension of the answer vector is high. Accordingly, extensive literature
has been dedicated to the question and the upper and lower bounds have been
matched up to constant factors [BUV18, SU17]. In this paper, we take a novel
approach to address this important optimality question. We first demonstrate an
intriguing central limit theorem phenomenon in the high-dimensional regime.
More precisely, we prove that a mechanism is approximately Gaussian
Differentially Private [DRS21] if the added noise satisfies certain conditions.
In particular, densities proportional to $\mathrm{e}^{-\|x\|_p^\alpha}$, where
$\|x\|_p$ is the standard $\ell_p$-norm, satisfies the conditions. Taking this
perspective, we make use of the Cramer--Rao inequality and show an "uncertainty
principle"-style result: the product of the privacy parameter and the
$\ell_2$-loss of the mechanism is lower bounded by the dimension. Furthermore,
the Gaussian mechanism achieves the constant-sharp optimal privacy-accuracy
trade-off among all such noises. Our findings are corroborated by numerical
experiments.
| stat |
Lifted samplers for partially ordered discrete state-spaces | A technique called lifting is employed in practice for avoiding that the
Markov chains simulated for sampling backtrack too often. It consists in
lifting the state-space to include direction variables for guiding these
chains. Its implementation is direct when the probability mass function
targeted is defined on a totally ordered set, such as that of a univariate
random variable taking values on the integers. In this paper, we adapt this
technique to the situation where only a partial order can be established and
explore its benefits. Important applications include simulation of systems
formed from binary variables, such as those described by the Ising model, and
variable selection when the marginal model probabilities can be evaluated, up
to a normalising constant. To accommodate for the situation where one does not
have access to these marginal model probabilities, a lifted trans-dimensional
sampler for partially ordered model spaces is introduced. We show through
theoretical analyses and empirical experiments that the lifted samplers
outperform their non-lifted counterparts in some situations, and this at no
extra computational cost. The code to reproduce all experiments is available
online.
| stat |
Spectral Subsampling MCMC for Stationary Time Series | Bayesian inference using Markov Chain Monte Carlo (MCMC) on large datasets
has developed rapidly in recent years. However, the underlying methods are
generally limited to relatively simple settings where the data have specific
forms of independence. We propose a novel technique for speeding up MCMC for
time series data by efficient data subsampling in the frequency domain. For
several challenging time series models, we demonstrate a speedup of up to two
orders of magnitude while incurring negligible bias compared to MCMC on the
full dataset. We also propose alternative control variates for variance
reduction based on data grouping and coreset constructions.
| stat |
Parameter Restrictions for the Sake of Identification: Is there Utility
in Asserting that Perhaps a Restriction Holds? | Statistical modeling can involve a tension between assumptions and
statistical identification. The law of the observable data may not uniquely
determine the value of a target parameter without invoking a key assumption,
and, while plausible, this assumption may not be obviously true in the
scientific context at hand. Moreover, there are many instances of key
assumptions which are untestable, hence we cannot rely on the data to resolve
the question of whether the target is legitimately identified. Working in the
Bayesian paradigm, we consider the grey zone of situations where a key
assumption, in the form of a parameter space restriction, is scientifically
reasonable but not incontrovertible for the problem being tackled.
Specifically, we investigate statistical properties that ensue if we structure
a prior distribution to assert that `maybe' or `perhaps' the assumption holds.
Technically this simply devolves to using a mixture prior distribution putting
just some prior weight on the assumption, or one of several assumptions,
holding. However, while the construct is straightforward, there is very little
literature discussing situations where Bayesian model averaging is employed
across a mix of fully identified and partially identified models.
| stat |
Independent Innovation Analysis for Nonlinear Vector Autoregressive
Process | The nonlinear vector autoregressive (NVAR) model provides an appealing
framework to analyze multivariate time series obtained from a nonlinear
dynamical system. However, the innovation (or error), which plays a key role by
driving the dynamics, is almost always assumed to be additive. Additivity
greatly limits the generality of the model, hindering analysis of general NVAR
processes which have nonlinear interactions between the innovations. Here, we
propose a new general framework called independent innovation analysis (IIA),
which estimates the innovations from completely general NVAR. We assume mutual
independence of the innovations as well as their modulation by an auxiliary
variable (which is often taken as the time index and simply interpreted as
nonstationarity). We show that IIA guarantees the identifiability of the
innovations with arbitrary nonlinearities, up to a permutation and
component-wise invertible nonlinearities. We also propose three estimation
frameworks depending on the type of the auxiliary variable. We thus provide the
first rigorous identifiability result for general NVAR, as well as very general
tools for learning such models.
| stat |
Computationally efficient surrogate-based optimization of coastal storm
waves heights and run-ups | Storm surges cause coastal inundations due to the setup of the water surface
resulting from atmospheric pressure, surface winds and breaking waves. The
latter is particularly difficult to be accounted for. For instance, it was
observed that during Typhoon Haiyan (2013, Philippines), a stretch of coral
reef near the coast, which was expected to protect the coastal communities,
actually amplified the waves. The propagation and breaking process of such
large nearshore waves can be successfully captured by a phase-resolving wave
model. Building defences requires estimating not only the maximum storm surge
height, but also maxima of breaking wave height and run-up on land, under
consideration of a variety of storm characteristics. However, the computational
complexity of the simulator makes optimization tasks impractical. To find the
maximum breaking wave (bore) height and the maximum run-up, we employ
optim-MICE, a new surrogate-based optimization scheme based on Gaussian Process
emulation and information theory. In two idealised settings, we efficiently
identify the conditions that create the largest storm waves at the coast using
a minimal number of simulations. This is the first surrogate-based optimization
of storm waves. It opens the door to previously inconceivable coastal risk
assessments.
| stat |
Causal Inference with Bipartite Designs | Bipartite experiments are a recent object of study in causal inference,
whereby treatment is applied to one set of units and outcomes of interest are
measured on a different set of units. These experiments are particularly useful
in settings where strong interference effects occur between units of a
bipartite graph. In market experiments for example, assigning treatment at the
seller-level and measuring outcomes at the buyer-level (or vice-versa) may lead
to causal models that better account for the interference that naturally occurs
between buyers and sellers. While bipartite experiments have been shown to
improve the estimation of causal effects in certain settings, the analysis must
be done carefully so as to not introduce unnecessary bias. We leverage the
generalized propensity score literature to show that we can obtain unbiased
estimates of causal effects for bipartite experiments under a standard set of
assumptions. We also discuss the construction of confidence sets with proper
coverage probabilities. We evaluate these methods using a bipartite graph from
a publicly available dataset studied in previous work on bipartite experiments,
showing through simulations a significant bias reduction and improved coverage.
| stat |
Combining Heterogeneous Spatial Datasets with Process-based Spatial
Fusion Models: A Unifying Framework | In modern spatial statistics, the structure of data that is collected has
become more heterogeneous. Depending on the type of spatial data, different
modeling strategies for spatial data are used. For example, a kriging approach
for geostatistical data; a Gaussian Markov random field model for lattice data;
or a log Gaussian Cox process for point-pattern data. Despite these different
modeling choices, the nature of underlying scientific data-generating (latent)
processes is often the same, which can be represented by some continuous
spatial surfaces. In this paper, we introduce a unifying framework for
process-based multivariate spatial fusion models. The framework can jointly
analyze all three aforementioned types of spatial data (or any combinations
thereof). Moreover, the framework accommodates different conditional
distributions for geostatistical and lattice data. We show that some
established approaches, such as linear models of coregionalization, can be
viewed as special cases of our proposed framework. We offer flexible and
scalable implementations in R using Stan and INLA. Simulation studies confirm
that the predictive performance of latent processes improves as we move from
univariate spatial models to multivariate spatial fusion models. The introduced
framework is illustrated using a cross-sectional study linked with a national
cohort dataset in Switzerland, we examine differences in underlying spatial
risk patterns between respiratory disease and lung cancer.
| stat |
Numerical tolerance for spectral decompositions of random matrices | We precisely quantify the impact of statistical error in the quality of a
numerical approximation to a random matrix eigendecomposition, and under mild
conditions, we use this to introduce an optimal numerical tolerance for
residual error in spectral decompositions of random matrices. We demonstrate
that terminating an eigendecomposition algorithm when the numerical error and
statistical error are of the same order results in computational savings with
no loss of accuracy. We also repair a flaw in a ubiquitous termination
condition, one in wide employ in several computational linear algebra
implementations. We illustrate the practical consequences of our stopping
criterion with an analysis of simulated and real networks. Our theoretical
results and real-data examples establish that the tradeoff between statistical
and numerical error is of significant import for data science.
| stat |
Computation-free Nonparametric testing for Local and Global Spatial
Autocorrelation with application to the Canadian Electorate | Measures of local and global spatial association are key tools for
exploratory spatial data analysis. Many such measures exist including Moran's
$I$, Geary's $C$, and the Getis-Ord $G$ and $G^*$ statistics. A parametric
approach to testing for significance relies on strong assumptions, which are
often not met by real world data. Alternatively, the most popular nonparametric
approach, the permutation test, imposes a large computational burden especially
for massive graphical networks. Hence, we propose a computation-free approach
to nonparametric permutation testing for local and global measures of spatial
autocorrelation stemming from generalizations of the Khintchine inequality from
functional analysis and the theory of $L^p$ spaces. Our methodology is
demonstrated on the results of the 2019 federal Canadian election in the
province of Alberta. We recorded the percentage of the vote gained by the
conservative candidate in each riding. This data is not normal, and the sample
size is fixed at $n=34$ ridings making the parametric approach invalid. In
contrast, running a classic permutation test for every riding, for multiple
test statistics, with various neighbourhood structures, and multiple testing
correction would require the simulation of millions of permutations. We are
able to achieve similar statistical power on this dataset to the permutation
test without the need for tedious simulation. We also consider data simulated
across the entire electoral map of Canada.
| stat |
Deep regularization and direct training of the inner layers of Neural
Networks with Kernel Flows | We introduce a new regularization method for Artificial Neural Networks
(ANNs) based on Kernel Flows (KFs). KFs were introduced as a method for kernel
selection in regression/kriging based on the minimization of the loss of
accuracy incurred by halving the number of interpolation points in random
batches of the dataset. Writing $f_\theta(x) = \big(f^{(n)}_{\theta_n}\circ
f^{(n-1)}_{\theta_{n-1}} \circ \dots \circ f^{(1)}_{\theta_1}\big)(x)$ for the
functional representation of compositional structure of the ANN, the inner
layers outputs $h^{(i)}(x) = \big(f^{(i)}_{\theta_i}\circ
f^{(i-1)}_{\theta_{i-1}} \circ \dots \circ f^{(1)}_{\theta_1}\big)(x)$ define a
hierarchy of feature maps and kernels $k^{(i)}(x,x')=\exp(- \gamma_i
\|h^{(i)}(x)-h^{(i)}(x')\|_2^2)$. When combined with a batch of the dataset
these kernels produce KF losses $e_2^{(i)}$ (the $L^2$ regression error
incurred by using a random half of the batch to predict the other half)
depending on parameters of inner layers $\theta_1,\ldots,\theta_i$ (and
$\gamma_i$). The proposed method simply consists in aggregating a subset of
these KF losses with a classical output loss. We test the proposed method on
CNNs and WRNs without alteration of structure nor output classifier and report
reduced test errors, decreased generalization gaps, and increased robustness to
distribution shift without significant increase in computational complexity. We
suspect that these results might be explained by the fact that while
conventional training only employs a linear functional (a generalized moment)
of the empirical distribution defined by the dataset and can be prone to
trapping in the Neural Tangent Kernel regime (under over-parameterizations),
the proposed loss function (defined as a nonlinear functional of the empirical
distribution) effectively trains the underlying kernel defined by the CNN
beyond regressing the data with that kernel.
| stat |
Hierarchical Bayesian Operational Modal Analysis: Theory and
Computations | This paper presents a hierarchical Bayesian modeling framework for the
uncertainty quantification in modal identification of linear dynamical systems
using multiple vibration data sets. This novel framework integrates the
state-of-the-art Bayesian formulations into a hierarchical setting aiming to
capture both the identification precision and the ensemble variability prompted
due to modeling errors. Such cutting-edge developments have been absent from
the modal identification literature, sustained as a long-standing problem at
the research spotlight. Central to this framework is a Gaussian hyper
probability model, whose mean and covariance matrix are unknown encapsulating
the uncertainty of the modal parameters. Detailed computation of this
hierarchical model is addressed under two major algorithms using Markov chain
Monte Carlo (MCMC) sampling and Laplace asymptotic approximation methods. Since
for a small number of data sets the hyper covariance matrix is often
unidentifiable, a practical remedy is suggested through the eigenbasis
transformation of the covariance matrix, which effectively reduces the number
of unknown hyper-parameters. It is also proved that under some conditions the
maximum a posteriori (MAP) estimation of the hyper mean and covariance coincide
with the ensemble mean and covariance computed using the MAP estimations
corresponding to multiple data sets. This interesting finding addresses
relevant concerns related to the outcome of the mainstream Bayesian methods in
capturing the stochastic variability from dissimilar data sets. Finally, the
dynamical response of a prototype structure tested on a shaking table subjected
to Gaussian white noise base excitation and the ambient vibration measurement
of a cable footbridge are employed to demonstrate the proposed framework.
| stat |
Multivariate Functional Additive Mixed Models | Multivariate functional data can be intrinsically multivariate like movement
trajectories in 2D or complementary like precipitation, temperature, and wind
speeds over time at a given weather station. We propose a multivariate
functional additive mixed model (multiFAMM) and show its application to both
data situations using examples from sports science (movement trajectories of
snooker players) and phonetic science (acoustic signals and articulation of
consonants). The approach includes linear and nonlinear covariate effects and
models the dependency structure between the dimensions of the responses using
multivariate functional principal component analysis. Multivariate functional
random intercepts capture both the auto-correlation within a given function and
cross-correlations between the multivariate functional dimensions. They also
allow us to model between-function correlations as induced by e.g.\ repeated
measurements or crossed study designs. Modeling the dependency structure
between the dimensions can generate additional insight into the properties of
the multivariate functional process, improves the estimation of random effects,
and yields corrected confidence bands for covariate effects. Extensive
simulation studies indicate that a multivariate modeling approach is more
parsimonious than fitting independent univariate models to the data while
maintaining or improving model fit.
| stat |
Doubly Debiased Lasso: High-Dimensional Inference under Hidden
Confounding | Inferring causal relationships or related associations from observational
data can be invalidated by the existence of hidden confounding. We focus on a
high-dimensional linear regression setting, where the measured covariates are
affected by hidden confounding and propose the {\em Doubly Debiased Lasso}
estimator for individual components of the regression coefficient vector. Our
advocated method simultaneously corrects both the bias due to estimation of
high-dimensional parameters as well as the bias caused by the hidden
confounding. We establish its asymptotic normality and also prove that it is
efficient in the Gauss-Markov sense. The validity of our methodology relies on
a dense confounding assumption, i.e. that every confounding variable affects
many covariates. The finite sample performance is illustrated with an extensive
simulation study and a genomic application.
| stat |
A novel Bayesian approach for variable selection in linear regression
models | We propose a novel Bayesian approach to the problem of variable selection in
multiple linear regression models. In particular, we present a hierarchical
setting which allows for direct specification of a-priori beliefs about the
number of nonzero regression coefficients as well as a specification of beliefs
that given coefficients are nonzero. To guarantee numerical stability, we adopt
a $g$-prior with an additional ridge parameter for the unknown regression
coefficients. In order to simulate from the joint posterior distribution an
intelligent random walk Metropolis-Hastings algorithm which is able to switch
between different models is proposed. Testing our algorithm on real and
simulated data illustrates that it performs at least on par and often even
better than other well-established methods. Finally, we prove that under some
nominal assumptions, the presented approach is consistent in terms of model
selection.
| stat |
The Landscape of R Packages for Automated Exploratory Data Analysis | The increasing availability of large but noisy data sets with a large number
of heterogeneous variables leads to the increasing interest in the automation
of common tasks for data analysis. The most time-consuming part of this process
is the Exploratory Data Analysis, crucial for better domain understanding, data
cleaning, data validation, and feature engineering.
There is a growing number of libraries that attempt to automate some of the
typical Exploratory Data Analysis tasks to make the search for new insights
easier and faster. In this paper, we present a systematic review of existing
tools for Automated Exploratory Data Analysis (autoEDA). We explore the
features of twelve popular R packages to identify the parts of analysis that
can be effectively automated with the current tools and to point out new
directions for further autoEDA development.
| stat |
Probabilistic forecast reconciliation under the Gaussian framework | Forecast reconciliation of multivariate time series is the process of mapping
a set of incoherent forecasts into coherent forecasts to satisfy a given set of
linear constraints. Commonly used projection matrix based approaches for point
forecast reconciliation are OLS (ordinary least squares), WLS (weighted least
squares), and MinT (minimum trace). Even though point forecast reconciliation
is a well-established field of research, the literature on generating
probabilistic forecasts subject to linear constraints is somewhat limited.
Available methods follow a two-step procedure. Firstly, it draws future sample
paths from the univariate models fitted to each series in the collection (which
are incoherent). Secondly, it uses a projection matrix based approach or
empirical copula based reordering approach to account for contemporaneous
correlations and linear constraints. The projection matrices are estimated
either by optimizing a scoring rule such as energy or variogram score, or
simply using a projection matrix derived for point forecast reconciliation.
This paper proves that (a) if the incoherent predictive distribution is
Gaussian then MinT minimizes the logarithmic scoring rule; and (b) the
logarithmic score of MinT for each marginal predictive density is smaller than
that of OLS. We show these theoretical results using a set of simulation
studies. We also evaluate them using the Australian domestic tourism data set.
| stat |
On the Banach spaces associated with multi-layer ReLU networks: Function
representation, approximation theory and gradient descent dynamics | We develop Banach spaces for ReLU neural networks of finite depth $L$ and
infinite width. The spaces contain all finite fully connected $L$-layer
networks and their $L^2$-limiting objects under bounds on the natural
path-norm. Under this norm, the unit ball in the space for $L$-layer networks
has low Rademacher complexity and thus favorable generalization properties.
Functions in these spaces can be approximated by multi-layer neural networks
with dimension-independent convergence rates.
The key to this work is a new way of representing functions in some form of
expectations, motivated by multi-layer neural networks. This representation
allows us to define a new class of continuous models for machine learning. We
show that the gradient flow defined this way is the natural continuous analog
of the gradient descent dynamics for the associated multi-layer neural
networks. We show that the path-norm increases at most polynomially under this
continuous gradient flow dynamics.
| stat |
A polynomial-time algorithm for learning nonparametric causal graphs | We establish finite-sample guarantees for a polynomial-time algorithm for
learning a nonlinear, nonparametric directed acyclic graphical (DAG) model from
data. The analysis is model-free and does not assume linearity, additivity,
independent noise, or faithfulness. Instead, we impose a condition on the
residual variances that is closely related to previous work on linear models
with equal variances. Compared to an optimal algorithm with oracle knowledge of
the variable ordering, the additional cost of the algorithm is linear in the
dimension $d$ and the number of samples $n$. Finally, we compare the proposed
algorithm to existing approaches in a simulation study.
| stat |
Neural Jump Ordinary Differential Equations: Consistent Continuous-Time
Prediction and Filtering | Combinations of neural ODEs with recurrent neural networks (RNN), like
GRU-ODE-Bayes or ODE-RNN are well suited to model irregularly observed time
series. While those models outperform existing discrete-time approaches, no
theoretical guarantees for their predictive capabilities are available.
Assuming that the irregularly-sampled time series data originates from a
continuous stochastic process, the $L^2$-optimal online prediction is the
conditional expectation given the currently available information. We introduce
the Neural Jump ODE (NJ-ODE) that provides a data-driven approach to learn,
continuously in time, the conditional expectation of a stochastic process. Our
approach models the conditional expectation between two observations with a
neural ODE and jumps whenever a new observation is made. We define a novel
training framework, which allows us to prove theoretical guarantees for the
first time. In particular, we show that the output of our model converges to
the $L^2$-optimal prediction. This can be interpreted as solution to a special
filtering problem. We provide experiments showing that the theoretical results
also hold empirically. Moreover, we experimentally show that our model
outperforms the baselines in more complex learning tasks and give comparisons
on real-world datasets.
| stat |
Active Learning for Decision-Making from Imbalanced Observational Data | Machine learning can help personalized decision support by learning models to
predict individual treatment effects (ITE). This work studies the reliability
of prediction-based decision-making in a task of deciding which action $a$ to
take for a target unit after observing its covariates $\tilde{x}$ and predicted
outcomes $\hat{p}(\tilde{y} \mid \tilde{x}, a)$. An example case is
personalized medicine and the decision of which treatment to give to a patient.
A common problem when learning these models from observational data is
imbalance, that is, difference in treated/control covariate distributions,
which is known to increase the upper bound of the expected ITE estimation
error. We propose to assess the decision-making reliability by estimating the
ITE model's Type S error rate, which is the probability of the model inferring
the sign of the treatment effect wrong. Furthermore, we use the estimated
reliability as a criterion for active learning, in order to collect new
(possibly expensive) observations, instead of making a forced choice based on
unreliable predictions. We demonstrate the effectiveness of this
decision-making aware active learning in two decision-making tasks: in
simulated data with binary outcomes and in a medical dataset with synthetic and
continuous treatment outcomes.
| stat |
Fusing Sufficient Dimension Reduction with Neural Networks | We consider the regression problem where the dependence of the response Y on
a set of predictors X is fully captured by the regression function E(Y |
X)=g(B'X), for an unknown function g and low rank parameter B matrix. We
combine neural networks with sufficient dimension reduction in order to remove
the limitation of small p and n of the latter. We show in simulations that the
proposed estimator is on par with competing sufficient dimension reduction
methods in small p and n settings, such as minimum average variance estimation
and conditional variance estimation. Among those, it is the only
computationally applicable in large p and n problems.
| stat |
Transport map accelerated adaptive importance sampling, and application
to inverse problems arising from multiscale stochastic reaction networks | In many applications, Bayesian inverse problems can give rise to probability
distributions which contain complexities due to the Hessian varying greatly
across parameter space. This complexity often manifests itself as lower
dimensional manifolds on which the likelihood function is invariant, or varies
very little. This can be due to trying to infer unobservable parameters, or due
to sloppiness in the model which is being used to describe the data. In such a
situation, standard sampling methods for characterising the posterior
distribution, which do not incorporate information about this structure, will
be highly inefficient.
In this paper, we seek to develop an approach to tackle this problem when
using adaptive importance sampling methods, by using optimal transport maps to
simplify posterior distributions which are concentrated on lower dimensional
manifolds. This approach is applicable to a whole range of problems for which
Monte Carlo Markov chain (MCMC) methods mix slowly.
We demonstrate the approach by considering inverse problems arising from
partially observed stochastic reaction networks. In particular, we consider
systems which exhibit multiscale behaviour, but for which only the slow
variables in the system are observable. We demonstrate that certain multiscale
approximations lead to more consistent approximations of the posterior than
others. The use of optimal transport maps stabilises the ensemble transform
adaptive importance sampling (ETAIS) method, and allows for efficient sampling
with smaller ensemble sizes. This approach allows us to take advantage of the
large increases of efficiency when using adaptive importance sampling methods
for previously intractable Bayesian inverse problems with complex posterior
structure.
| stat |
Linear regression and its inference on noisy network-linked data | Linear regression on a set of observations linked by a network has been an
essential tool in modeling the relationship between response and covariates
with additional network data. Despite its wide range of applications in many
areas, such as in social sciences and health-related research, the problem has
not been well-studied in statistics so far. Previous methods either lack
inference tools or rely on restrictive assumptions on social effects and
usually assume that networks are observed without errors, which is unrealistic
in many problems. This paper proposes a linear regression model with
nonparametric network effects. The model does not assume that the relational
data or network structure is exactly observed; thus, the method can be provably
robust to a certain network perturbation level. A set of asymptotic inference
results is established under a general requirement of the network observational
errors, and the robustness of this method is studied in the specific setting
when the errors come from random network models. We discover a phase-transition
phenomenon of the inference validity concerning the network density when no
prior knowledge of the network model is available while also showing a
significant improvement achieved by knowing the network model. As a by-product
of this analysis, we derive a rate-optimal concentration bound for random
subspace projection that may be of independent interest. Extensive simulation
studies are conducted to verify these theoretical results and demonstrate the
advantage of the proposed method over existing work in terms of accuracy and
computational efficiency under different data-generating models. The method is
then applied to adolescent network data to study the gender and racial
difference in social activities.
| stat |
Boosting insights in insurance tariff plans with tree-based machine
learning methods | Pricing actuaries typically operate within the framework of generalized
linear models (GLMs). With the upswing of data analytics, our study puts focus
on machine learning methods to develop full tariff plans built from both the
frequency and severity of claims. We adapt the loss functions used in the
algorithms such that the specific characteristics of insurance data are
carefully incorporated: highly unbalanced count data with excess zeros and
varying exposure on the frequency side combined with scarce, but potentially
long-tailed data on the severity side. A key requirement is the need for
transparent and interpretable pricing models which are easily explainable to
all stakeholders. We therefore focus on machine learning with decision trees:
starting from simple regression trees, we work towards more advanced ensembles
such as random forests and boosted trees. We show how to choose the optimal
tuning parameters for these models in an elaborate cross-validation scheme, we
present visualization tools to obtain insights from the resulting models and
the economic value of these new modeling approaches is evaluated. Boosted trees
outperform the classical GLMs, allowing the insurer to form profitable
portfolios and to guard against potential adverse risk selection.
| stat |
Boosting for Comparison-Based Learning | We consider the problem of classification in a comparison-based setting:
given a set of objects, we only have access to triplet comparisons of the form
"object $x_i$ is closer to object $x_j$ than to object $x_k$." In this paper we
introduce TripletBoost, a new method that can learn a classifier just from such
triplet comparisons. The main idea is to aggregate the triplets information
into weak classifiers, which can subsequently be boosted to a strong
classifier. Our method has two main advantages: (i) it is applicable to data
from any metric space, and (ii) it can deal with large scale problems using
only passively obtained and noisy triplets. We derive theoretical
generalization guarantees and a lower bound on the number of necessary
triplets, and we empirically show that our method is both competitive with
state of the art approaches and resistant to noise.
| stat |
A sampling criterion for constrained Bayesian optimization with
uncertainties | We consider the problem of chance constrained optimization where it is sought
to optimize a function and satisfy constraints, both of which are affected by
uncertainties. The real world declinations of this problem are particularly
challenging because of their inherent computational cost.
To tackle such problems, we propose a new Bayesian optimization method. It
applies to the situation where the uncertainty comes from some of the inputs,
so that it becomes possible to define an acquisition criterion in the joint
controlled-uncontrolled input space. The main contribution of this work is an
acquisition criterion that accounts for both the average improvement in
objective function and the constraint reliability. The criterion is derived
following the Stepwise Uncertainty Reduction logic and its maximization
provides both optimal controlled and uncontrolled parameters. Analytical
expressions are given to efficiently calculate the criterion. Numerical studies
on test functions are presented. It is found through experimental comparisons
with alternative sampling criteria that the adequation between the sampling
criterion and the problem contributes to the efficiency of the overall
optimization. As a side result, an expression for the variance of the
improvement is given.
| stat |
Self Organizing Supply Chains for Micro-Prediction: Present and Future
uses of the ROAR Protocol | A multi-agent system is trialed as a means of crowd-sourcing inexpensive but
high quality streams of predictions. Each agent is a microservice embodying
statistical models and endowed with economic self-interest. The ability to fork
and modify simple agents is granted to a large number of employees in a firm
and empirical lessons are reported. We suggest that one plausible trajectory
for this project is the creation of a Prediction Web.
| stat |
A novel estimand to adjust for rescue treatment in clinical trials | The interpretation of randomised clinical trial results is often complicated
by intercurrent events. For instance, rescue medication is sometimes given to
patients in response to worsening of their disease, either in addition to the
randomised treatment or in its place. The use of such medication complicates
the interpretation of the intention-to-treat analysis. In view of this, we
propose a novel estimand defined as the intention-to-treat effect that would
have been observed, had patients on the active arm been switched to rescue
medication if and only if they would have been switched when randomised to
control. This enables us to disentangle the treatment effect from the effect of
rescue medication on a patient's outcome, while avoiding the strong
extrapolations that are typically needed when inferring what the
intention-to-treat effect would have been in the absence of rescue medication.
We develop an inverse probability weighting method to estimate this estimand
under specific untestable assumptions, in view of which we propose a
sensitivity analysis. We use the method for the analysis of a clinical trial
conducted by Janssen Pharmaceuticals, in which chronically ill patients can
switch to rescue medication for ethical reasons. Monte Carlo simulations
confirm that the proposed estimator is unbiased in moderate sample sizes.
| stat |
Gaussian Process Priors for Dynamic Paired Comparison Modelling | Dynamic paired comparison models, such as Elo and Glicko, are frequently used
for sports prediction and ranking players or teams. We present an alternative
dynamic paired comparison model which uses a Gaussian Process (GP) as a prior
for the time dynamics rather than the Markovian dynamics usually assumed. In
addition, we show that the GP model can easily incorporate covariates. We
derive an efficient approximate Bayesian inference procedure based on the
Laplace Approximation and sparse linear algebra. We select hyperparameters by
maximising their marginal likelihood using Bayesian Optimisation, comparing the
results against random search. Finally, we fit and evaluate the model on the
2018 season of ATP tennis matches, where it performs competitively,
outperforming Elo and Glicko on log loss, particularly when surface covariates
are included.
| stat |
Ascent with Quadratic Assistance for the Construction of Exact
Experimental Designs | In the area of statistical planning, there is a large body of theoretical
knowledge and computational experience concerning so-called optimal approximate
designs of experiments. However, for an approximate design to be executed in
practice, it must be converted into an exact, i.e., integer, design, which is
usually done via rounding procedures. Although rapid, rounding procedures have
many drawbacks; in particular, they often yield worse exact designs than
heuristics that do not require approximate designs at all.
In this paper, we build on an alternative principle of utilizing optimal
approximate designs for the computation of optimal, or nearly-optimal, exact
designs. The principle, which we call ascent with quadratic assistance (AQuA),
is an integer programming method based on the quadratic approximation of the
design criterion in the neighborhood of the optimal approximate information
matrix.
To this end, we present quadratic approximations of all Kiefer's criteria
with an integer parameter, including D- and A-optimality and, by a model
transformation, I-optimality. Importantly, we prove a low-rank property of the
associated quadratic forms, which enables us to apply AQuA to large design
spaces, for example via mixed integer conic quadratic solvers. We numerically
demonstrate the robustness and superior performance of the proposed method for
models under various types of constraints. More precisely, we compute optimal
size-constrained exact designs for the model of spring-balance weighing, and
optimal symmetric marginally restricted exact designs for the Scheffe mixture
model. We also show how can iterative application of AQuA be used for a
stratified information-based subsampling of large datasets under a lower bound
on the quality and an upper bound on the cost of the subsample.
| stat |
Bayesian Optimization of Hyperparameters when the Marginal Likelihood is
Estimated by MCMC | Bayesian models often involve a small set of hyperparameters determined by
maximizing the marginal likelihood. Bayesian optimization is a popular
iterative method where a Gaussian process posterior of the underlying function
is sequentially updated by new function evaluations. An acquisition strategy
uses this posterior distribution to decide where to place the next function
evaluation. We propose a novel Bayesian optimization framework for situations
where the user controls the computational effort, and therefore the precision
of the function evaluations. This is a common situation in econometrics where
the marginal likelihood is often computed by Markov Chain Monte Carlo (MCMC)
methods, with the precision of the marginal likelihood estimate determined by
the number of MCMC draws. The proposed acquisition strategy gives the optimizer
the option to explore the function with cheap noisy evaluations and therefore
finds the optimum faster. Prior hyperparameter estimation in the steady-state
Bayesian vector autoregressive (BVAR) model on US macroeconomic time series
data is used for illustration. The proposed method is shown to find the optimum
much quicker than traditional Bayesian optimization or grid search.
| stat |
On approximating $\nabla f$ with neural networks | Consider a feedforward neural network $\psi: \mathbb{R}^d\rightarrow
\mathbb{R}^d$ such that $\psi\approx \nabla f$, where $f:\mathbb{R}^d
\rightarrow \mathbb{R}$ is a smooth function, therefore $\psi$ must satisfy
$\partial_j \psi_i = \partial_i \psi_j$ pointwise. We prove a theorem that a
$\psi$ network with more than one hidden layer can only represent one feature
in its first hidden layer; this is a dramatic departure from the well-known
results for one hidden layer. The proof of the theorem is straightforward,
where two backward paths and a weight-tying matrix play the key roles. We then
present the alternative, the implicit parametrization, where the neural network
is $\phi: \mathbb{R}^d \rightarrow \mathbb{R}$ and $\nabla \phi \approx \nabla
f$; in addition, a "soft analysis" of $\nabla \phi$ gives a dual perspective on
the theorem. Throughout, we come back to recent probabilistic models that are
formulated as $\nabla \phi \approx \nabla f$, and conclude with a critique of
denoising autoencoders.
| stat |
Unsupervised Recalibration | Unsupervised recalibration (URC) is a general way to improve the accuracy of
an already trained probabilistic classification or regression model upon
encountering new data while deployed in the field. URC does not require any
ground truth associated with the new field data. URC merely observes the
model's predictions and recognizes when the training set is not representative
of field data, and then corrects to remove any introduced bias.
URC can be particularly useful when applied separately to different
subpopulations observed in the field that were not considered as features when
training the machine learning model. This makes it possible to exploit
subpopulation information without retraining the model or even having ground
truth for some or all subpopulations available.
Additionally, if these subpopulations are the object of study, URC serves to
determine the correct ground truth distributions for them, where naive
aggregation methods, like averaging the model's predictions, systematically
underestimate their differences.
| stat |
Fast Approximation and Estimation Bounds of Kernel Quadrature for
Infinitely Wide Models | An infinitely wide model is a weighted integration $\int \varphi(x,v) d
\mu(v)$ of feature maps. This model excels at handling an infinite number of
features, and thus it has been adopted to the theoretical study of deep
learning. Kernel quadrature is a kernel-based numerical integration scheme
developed for fast approximation of expectations $\int f(x) d p(x)$. In this
study, regarding the weight $\mu$ as a signed (or complex/vector-valued)
distribution of parameters, we develop the general kernel quadrature (GKQ) for
parameter distributions. The proposed method can achieve a fast approximation
rate $O(e^{-p})$ with parameter number $p$, which is faster than the
traditional Barron's rate, and a fast estimation rate $\widetilde{O}(1/n)$ with
sample size $n$. As a result, we have obtained a new norm-based complexity
measure for infinitely wide models. Since the GKQ implicitly conducts the
empirical risk minimization, we can understand that the complexity measure also
reflects the generalization performance in the gradient learning setup.
| stat |
Transfer Entropy Rate Through Lempel-Ziv Complexity | In this article we present a methodology to estimate the Transfer Entropy
Rate between two systems through the Lempel-Ziv complexity. This methodology
carries a set of practical advantages: it can be estimated from two single
discrete series of measures, it is not computationally expensive and it does
not assume any model for the data. The results of simulations over three
different unidirectional coupled systems, suggest that this methodology can be
used to assess the direction and strength of the information flow between
systems.
| stat |
Prescriptive Cluster-Dependent Support Vector Machines with an
Application to Reducing Hospital Readmissions | We augment linear Support Vector Machine (SVM) classifiers by adding three
important features: (i) we introduce a regularization constraint to induce a
sparse classifier; (ii) we devise a method that partitions the positive class
into clusters and selects a sparse SVM classifier for each cluster; and (iii)
we develop a method to optimize the values of controllable variables in order
to reduce the number of data points which are predicted to have an undesirable
outcome, which, in our setting, coincides with being in the positive class. The
latter feature leads to personalized prescriptions/recommendations. We apply
our methods to the problem of predicting and preventing hospital readmissions
within 30-days from discharge for patients that underwent a general surgical
procedure. To that end, we leverage a large dataset containing over 2.28
million patients who had surgeries in the period 2011--2014 in the U.S. The
dataset has been collected as part of the American College of Surgeons National
Surgical Quality Improvement Program (NSQIP).
| stat |
Pitfalls to Avoid when Interpreting Machine Learning Models | Modern requirements for machine learning (ML) models include both high
predictive performance and model interpretability. A growing number of
techniques provide model interpretations, but can lead to wrong conclusions if
applied incorrectly. We illustrate pitfalls of ML model interpretation such as
bad model generalization, dependent features, feature interactions or
unjustified causal interpretations. Our paper addresses ML practitioners by
raising awareness of pitfalls and pointing out solutions for correct model
interpretation, as well as ML researchers by discussing open issues for further
research.
| stat |
Learning Latent and Hierarchical Structures in Cognitive Diagnosis
Models | Cognitive Diagnosis Models (CDMs) are a special family of discrete latent
variable models that are widely used in modern educational, psychological,
social and biological sciences. A key component of CDMs is a binary $Q$-matrix
characterizing the dependence structure between the items and the latent
attributes. Additionally, researchers also assume in many applications certain
hierarchical structures among the latent attributes to characterize their
dependence. In most CDM applications, the attribute-attribute hierarchical
structures, the item-attribute $Q$-matrix, the item-level diagnostic model, as
well as the number of latent attributes, need to be fully or partially
pre-specified, which however may be subjective and misspecified as noted by
many recent studies. This paper considers the problem of jointly learning these
latent and hierarchical structures in CDMs from observed data with minimal
model assumptions. Specifically, a penalized likelihood approach is proposed to
select the number of attributes and estimate the latent and hierarchical
structures simultaneously. An efficient expectation-maximization (EM) algorithm
and a latent structure recovery algorithm are developed, and statistical
consistency theory is also established under mild conditions. The good
performance of the proposed method is illustrated by simulation studies and a
real data application in educational assessment.
| stat |
The Concordance coefficient: An alternative to the Kruskal-Wallis test | Kendall rank correlation coefficient is used to measure the ordinal
association between two measurements. In this paper, we introduce the
Concordance coefficient as a generalization of the Kendall rank correlation,
and illustrate its use to measure the ordinal association between quantity and
quality measures when two or more samples are considered. In this sense, the
Concordance coefficient can be seen as a generalization of the Kendall rank
correlation coefficient and an alternative to the non-parametric mean
rank-based methods to compare two or more samples. A comparison of the proposed
Concordance coefficient and the classical Kruskal-Wallis statistic is presented
through a comparison of exact distributions of both statistics.
| stat |
An efficient and accurate approximation to the distribution of quadratic
forms of Gaussian variables | In computational and applied statistics, it is of great interest to get fast
and accurate calculation for the distributions of the quadratic forms of
Gaussian random variables. This paper presents a novel approximation strategy
that contains two developments. First, we propose a faster numerical procedure
in computing the moments of the quadratic forms. Second, we establish a general
moment-matching framework for distribution approximation, which covers existing
approximation methods for the distributions of the quadratic forms of Gaussian
variables. Under this framework, a novel moment-ratio method (MR) is proposed
to match the ratio of skewness and kurtosis based on the gamma distribution.
Our extensive simulations show that 1) MR is almost as accurate as the exact
distribution calculation and is much more efficient; 2) comparing with existing
approximation methods, MR significantly improves the accuracy of approximating
far right tail probabilities. The proposed method has wide applications. For
example, it is a better choice than existing methods for facilitating
hypothesis testing in big data analysis, where efficient and accurate
calculation of very small $p$-values is desired. An R package Qapprox that
implements related methods is available on CRAN.
| stat |
Applying economic measures to lapse risk management with machine
learning approaches | Modeling policyholders lapse behaviors is important to a life insurer since
lapses affect pricing, reserving, profitability, liquidity, risk management, as
well as the solvency of the insurer. Lapse risk is indeed the most significant
life underwriting risk according to European Insurance and Occupational
Pensions Authority's Quantitative Impact Study QIS5. In this paper, we
introduce two advanced machine learning algorithms for lapse modeling. Then we
evaluate the performance of different algorithms by means of classical
statistical accuracy and profitability measure. Moreover, we adopt an
innovative point of view on the lapse prediction problem that comes from churn
management. We transform the classification problem into a regression question
and then perform optimization, which is new for lapse risk management. We apply
different algorithms to a large real-world insurance dataset. Our results show
that XGBoost and SVM outperform CART and logistic regression, especially in
terms of the economic validation metric. The optimization after transformation
brings out significant and consistent increases in economic gains.
| stat |
Efficient data augmentation using graph imputation neural networks | Recently, data augmentation in the semi-supervised regime, where unlabeled
data vastly outnumbers labeled data, has received a considerable attention. In
this paper, we describe an efficient technique for this task, exploiting a
recent framework we proposed for missing data imputation called graph
imputation neural network (GINN). The key idea is to leverage both supervised
and unsupervised data to build a graph of similarities between points in the
dataset. Then, we augment the dataset by severely damaging a few of the nodes
(up to 80\% of their features), and reconstructing them using a variation of
GINN. On several benchmark datasets, we show that our method can obtain
significant improvements compared to a fully-supervised model, and we are able
to augment the datasets up to a factor of 10x. This points to the power of
graph-based neural networks to represent structural affinities in the samples
for tasks of data reconstruction and augmentation.
| stat |
Sparse (group) learning with Lipschitz loss functions: a unified
analysis | We study a family of sparse estimators defined as minimizers of some
empirical Lipschitz loss function---which include hinge, logistic and quantile
regression losses---with a convex, sparse or group-sparse regularization. In
particular, we consider the L1-norm on the coefficients, its sorted Slope
version, and the Group L1-L2 extension. First, we propose a theoretical
framework which simultaneously derives new L2 estimation upper bounds for all
three regularization schemes. For L1 and Slope regularizations, our bounds
scale as $(k^*/n) \log(p/k^*)$---$n\times p$ is the size of the design matrix
and $k^*$ the dimension of the theoretical loss minimizer $\beta^*$---matching
the optimal minimax rate achieved for the least-squares case. For Group L1-L2
regularization, our bounds scale as $(s^*/n) \log\left( G / s^* \right) + m^* /
n$---$G$ is the total number of groups and $m^*$ the number of coefficients in
the $s^*$ groups which contain $\beta^*$---and improve over the least-squares
case. We additionally show that when the signal is strongly group-sparse Group
L1-L2 is superior to L1 and Slope. Our bounds are achieved both in probability
and in expectation, under common assumptions in the literature. Second, we
propose an accelerated proximal algorithm which computes the convex estimators
studied when the number of variables is of the order of $100,000$. We
additionally compare their statistical performance of our estimators against
standard baselines for settings where the signal is either sparse or
group-sparse. Our experiments findings reveal (i) the good empirical
performance of L1 and Slope regularizations for sparse binary classification
problems, (ii) the superiority of Group L1-L2 regularization for group-sparse
classification problems and (iii) the appealing properties of sparse quantile
regression estimators for sparse regression problems with heteroscedastic
noise.
| stat |
Statistical Inference and Power Analysis for Direct and Spillover
Effects in Two-Stage Randomized Experiments | Two-stage randomized experiments are becoming an increasingly popular
experimental design for causal inference when the outcome of one unit may be
affected by the treatment assignments of other units in the same cluster. In
this paper, we provide a methodological framework for general tools of
statistical inference and power analysis for two-stage randomized experiments.
Under the randomization-based framework, we propose unbiased point estimators
of direct and spillover effects, construct conservative variance estimators,
develop hypothesis testing procedures, and derive sample size formulas. We also
establish the equivalence relationships between the randomization-based and
regression based methods. We theoretically compare the two-stage randomized
design with the completely randomized and cluster randomized designs, which
represent two limiting designs. Finally, we conduct simulation studies to
evaluate the empirical performance of our sample size formulas. For empirical
illustration, the proposed methodology is applied to the analysis of the data
from a field experiment on a job placement assistance program.
| stat |
A One-Class Classification Decision Tree Based on Kernel Density
Estimation | One-class Classification (OCC) is an area of machine learning which addresses
prediction based on unbalanced datasets. Basically, OCC algorithms achieve
training by means of a single class sample, with potentially some additional
counter-examples. The current OCC models give satisfaction in terms of
performance, but there is an increasing need for the development of
interpretable models. In the present work, we propose a one-class model which
addresses concerns of both performance and interpretability. Our hybrid OCC
method relies on density estimation as part of a tree-based learning algorithm,
called One-Class decision Tree (OC-Tree). Within a greedy and recursive
approach, our proposal rests on kernel density estimation to split a data
subset on the basis of one or several intervals of interest. Thus, the OC-Tree
encloses data within hyper-rectangles of interest which can be described by a
set of rules. Against state-of-the-art methods such as Cluster Support Vector
Data Description (ClusterSVDD), One-Class Support Vector Machine (OCSVM) and
isolation Forest (iForest), the OC-Tree performs favorably on a range of
benchmark datasets. Furthermore, we propose a real medical application for
which the OC-Tree has demonstrated its effectiveness, through the ability to
tackle interpretable diagnosis aid based on unbalanced datasets.
| stat |
Learning in Volatile Environments with the Bayes Factor Surprise | Surprise-based learning allows agents to rapidly adapt to non-stationary
stochastic environments characterized by sudden changes. We show that exact
Bayesian inference in a hierarchical model gives rise to a surprise-modulated
trade-off between forgetting old observations and integrating them with the new
ones. The modulation depends on a probability ratio, which we call "Bayes
Factor Surprise", that tests the prior belief against the current belief. We
demonstrate that in several existing approximate algorithms the Bayes Factor
Surprise modulates the rate of adaptation to new observations. We derive three
novel surprised-based algorithms, one in the family of particle filters, one in
the family of variational learning, and the other in the family of message
passing, that have constant scaling in observation sequence length and
particularly simple update dynamics for any distribution in the exponential
family. Empirical results show that these surprise-based algorithms estimate
parameters better than alternative approximate approaches and reach levels of
performance comparable to computationally more expensive algorithms. The Bayes
Factor Surprise is related to but different from Shannon Surprise. In two
hypothetical experiments, we make testable predictions for physiological
indicators that dissociate the Bayes Factor Surprise from Shannon Surprise. The
theoretical insight of casting various approaches as surprise-based learning,
as well as the proposed online algorithms, may be applied to the analysis of
animal and human behavior, and to reinforcement learning in non-stationary
environments.
| stat |
Choosing the correlation structure of mixed effect models for
experiments with stimuli | The design of experiments in psychology can often be summarized to
participants reacting to stimuli. For such an experiment, the mixed effects
model with crossed random effects is usually the appropriate tool to analyse
the data because it considers the sampling of both participants and stimuli.
However, these models let to users several choices when analysing data and this
practice may be disruptive for researchers trained to a set of standardized
analysis such as ANOVA. In the present article, we are focusing on the choice
of the correlation structure of the data, because it is both subtle and
influential on the results of the analysis. We provide an overview of several
correlation structures used in the literature and we propose a new one that is
the natural extension of the repeated measures ANOVA. A large simulation study
shows that correlation structures that are either too simple or too complex
fail to deliver credible results, even for designs with only three variables.
We also show how the design of the experiment influences the correlation
structure of the data. Moreover, we provide R code to estimate all the
correlation structures presented in this article, as well as functions
implemented in an R package to compute our new proposal.
| stat |
Limiting Bias from Test-Control Interference in Online Marketplace
Experiments | In an A/B test, the typical objective is to measure the total average
treatment effect (TATE), which measures the difference between the average
outcome if all users were treated and the average outcome if all users were
untreated. However, a simple difference-in-means estimator will give a biased
estimate of the TATE when outcomes of control units depend on the outcomes of
treatment units, an issue we refer to as test-control interference. Using a
simulation built on top of data from Airbnb, this paper considers the use of
methods from the network interference literature for online marketplace
experimentation. We model the marketplace as a network in which an edge exists
between two sellers if their goods substitute for one another. We then simulate
seller outcomes, specifically considering a "status quo" context and
"treatment" context that forces all sellers to lower their prices. We use the
same simulation framework to approximate TATE distributions produced by using
blocked graph cluster randomization, exposure modeling, and the Hajek estimator
for the difference in means. We find that while blocked graph cluster
randomization reduces the bias of the naive difference-in-means estimator by as
much as 62%, it also significantly increases the variance of the estimator. On
the other hand, the use of more sophisticated estimators produces mixed
results. While some provide (small) additional reductions in bias and small
reductions in variance, others lead to increased bias and variance. Overall,
our results suggest that experiment design and analysis techniques from the
network experimentation literature are promising tools for reducing bias due to
test-control interference in marketplace experiments.
| stat |
Distributional Robustness with IPMs and links to Regularization and GANs | Robustness to adversarial attacks is an important concern due to the
fragility of deep neural networks to small perturbations and has received an
abundance of attention in recent years. Distributionally Robust Optimization
(DRO), a particularly promising way of addressing this challenge, studies
robustness via divergence-based uncertainty sets and has provided valuable
insights into robustification strategies such as regularization. In the context
of machine learning, the majority of existing results have chosen
$f$-divergences, Wasserstein distances and more recently, the Maximum Mean
Discrepancy (MMD) to construct uncertainty sets. We extend this line of work
for the purposes of understanding robustness via regularization by studying
uncertainty sets constructed with Integral Probability Metrics (IPMs) - a large
family of divergences including the MMD, Total Variation and Wasserstein
distances. Our main result shows that DRO under \textit{any} choice of IPM
corresponds to a family of regularization penalties, which recover and improve
upon existing results in the setting of MMD and Wasserstein distances. Due to
the generality of our result, we show that other choices of IPMs correspond to
other commonly used penalties in machine learning. Furthermore, we extend our
results to shed light on adversarial generative modelling via $f$-GANs,
constituting the first study of distributional robustness for the $f$-GAN
objective. Our results unveil the inductive properties of the discriminator set
with regards to robustness, allowing us to give positive comments for several
penalty-based GAN methods such as Wasserstein-, MMD- and Sobolev-GANs. In
summary, our results intimately link GANs to distributional robustness, extend
previous results on DRO and contribute to our understanding of the link between
regularization and robustness at large.
| stat |
PermuteAttack: Counterfactual Explanation of Machine Learning Credit
Scorecards | This paper is a note on new directions and methodologies for validation and
explanation of Machine Learning (ML) models employed for retail credit scoring
in finance. Our proposed framework draws motivation from the field of
Artificial Intelligence (AI) security and adversarial ML where the need for
certifying the performance of the ML algorithms in the face of their
overwhelming complexity poses a need for rethinking the traditional notions of
model architecture selection, sensitivity analysis and stress testing. Our
point of view is that the phenomenon of adversarial perturbations when detached
from the AI security domain, has purely algorithmic roots and fall within the
scope of model risk assessment. We propose a model criticism and explanation
framework based on adversarially generated counterfactual examples for tabular
data. A counterfactual example to a given instance in this context is defined
as a synthetically generated data point sampled from the estimated data
distribution which is treated differently by a model. The counterfactual
examples can be used to provide a black-box instance-level explanation of the
model behaviour as well as studying the regions in the input space where the
model performance deteriorates. Adversarial example generating algorithms are
extensively studied in the image and natural language processing (NLP) domains.
However, most financial data come in tabular format and naive application of
the existing techniques on this class of datasets generates unrealistic
samples. In this paper, we propose a counterfactual example generation method
capable of handling tabular data including discrete and categorical variables.
Our proposed algorithm uses a gradient-free optimization based on genetic
algorithms and therefore is applicable to any classification model.
| stat |
Computational Phenotype Discovery via Probabilistic Independence | Computational Phenotype Discovery research has taken various pragmatic
approaches to disentangling phenotypes from the episodic observations in
Electronic Health Records. In this work, we use transformation into continuous,
longitudinal curves to abstract away the sparse irregularity of the data, and
we introduce probabilistic independence as a guiding principle for
disentangling phenotypes into patterns that may more closely match true
pathophysiologic mechanisms. We use the identification of liver disease
patterns that presage development of Hepatocellular Carcinoma as a
proof-of-concept demonstration.
| stat |
Bootstrapping spectral statistics in high dimensions | Statistics derived from the eigenvalues of sample covariance matrices are
called spectral statistics, and they play a central role in multivariate
testing. Although bootstrap methods are an established approach to
approximating the laws of spectral statistics in low-dimensional problems,
these methods are relatively unexplored in the high-dimensional setting. The
aim of this paper is to focus on linear spectral statistics as a class of
prototypes for developing a new bootstrap in high-dimensions --- and we refer
to this method as the Spectral Bootstrap. In essence, the method originates
from the parametric bootstrap, and is motivated by the notion that, in high
dimensions, it is difficult to obtain a non-parametric approximation to the
full data-generating distribution. From a practical standpoint, the method is
easy to use, and allows the user to circumvent the difficulties of complex
asymptotic formulas for linear spectral statistics. In addition to proving the
consistency of the proposed method, we provide encouraging empirical results in
a variety of settings. Lastly, and perhaps most interestingly, we show through
simulations that the method can be applied successfully to statistics outside
the class of linear spectral statistics, such as the largest sample eigenvalue
and others.
| stat |
Approximate inferences for nonlinear mixed effects models with scale
mixtures of skew-normal distributions | Nonlinear mixed effects models have received a great deal of attention in the
statistical literature in recent years because of their flexibility in handling
longitudinal studies, including human immunodeficiency virus viral dynamics,
pharmacokinetic analyses, and studies of growth and decay. A standard
assumption in nonlinear mixed effects models for continuous responses is that
the random effects and the within-subject errors are normally distributed,
making the model sensitive to outliers. We present a novel class of asymmetric
nonlinear mixed effects models that provides efficient parameters estimation in
the analysis of longitudinal data. We assume that, marginally, the random
effects follow a multivariate scale mixtures of skew--normal distribution and
that the random errors follow a symmetric scale mixtures of normal
distribution, providing an appealing robust alternative to the usual normal
distribution. We propose an approximate method for maximum likelihood
estimation based on an EM-type algorithm that produces approximate maximum
likelihood estimates and significantly reduces the numerical difficulties
associated with the exact maximum likelihood estimation. Techniques for
prediction of future responses under this class of distributions are also
briefly discussed. The methodology is illustrated through an application to
Theophylline kinetics data and through some simulating studies.
| stat |
Multi-Product Dynamic Pricing in High-Dimensions with Heterogeneous
Price Sensitivity | We consider the problem of multi-product dynamic pricing, in a contextual
setting, for a seller of differentiated products. In this environment, the
customers arrive over time and products are described by high-dimensional
feature vectors. Each customer chooses a product according to the widely used
Multinomial Logit (MNL) choice model and her utility depends on the product
features as well as the prices offered. The seller a-priori does not know the
parameters of the choice model but can learn them through interactions with
customers. The seller's goal is to design a pricing policy that maximizes her
cumulative revenue. This model is motivated by online marketplaces such as
Airbnb platform and online advertising. We measure the performance of a pricing
policy in terms of regret, which is the expected revenue loss with respect to a
clairvoyant policy that knows the parameters of the choice model in advance and
always sets the revenue-maximizing prices. We propose a pricing policy, named
M3P, that achieves a $T$-period regret of $O(\log(Td) ( \sqrt{T}+ d\log(T)))$
under heterogeneous price sensitivity for products with features of dimension
$d$. We also use tools from information theory to prove that no policy can
achieve worst-case $T$-regret better than $\Omega(\sqrt{T})$.
| stat |
Multiple Causes: A Causal Graphical View | Unobserved confounding is a major hurdle for causal inference from
observational data. Confounders---the variables that affect both the causes and
the outcome---induce spurious non-causal correlations between the two. Wang &
Blei (2018) lower this hurdle with "the blessings of multiple causes," where
the correlation structure of multiple causes provides indirect evidence for
unobserved confounding. They leverage these blessings with an algorithm, called
the deconfounder, that uses probabilistic factor models to correct for the
confounders. In this paper, we take a causal graphical view of the
deconfounder. In a graph that encodes shared confounding, we show how the
multiplicity of causes can help identify intervention distributions. We then
justify the deconfounder, showing that it makes valid inferences of the
intervention. Finally, we expand the class of graphs, and its theory, to those
that include other confounders and selection variables. Our results expand the
theory in Wang & Blei (2018), justify the deconfounder for causal graphs, and
extend the settings where it can be used.
| stat |
Model assessment for time series dynamics using copula spectral
densities: a graphical tool | Finding parametric models that accurately describe the dependence structure
of observed data is a central task in the analysis of time series. Classical
frequency domain methods provide a popular set of tools for fitting and
diagnostics of time series models, but their applicability is seriously
impacted by the limitations of covariances as a measure of dependence.
Motivated by recent developments of frequency domain methods that are based on
copulas instead of covariances, we propose a novel graphical tool that allows
to access the quality of time series models for describing dependencies that go
beyond linearity. We provide a thorough theoretical justification of our
approach and show in simulations that it can successfully distinguish between
subtle differences of time series dynamics, including non-linear dynamics which
result from GARCH and EGARCH models. We also demonstrate the utility of the
proposed tools through an application to modeling returns of the S&P 500 stock
market index.
| stat |
Revisiting the orbital tracking problem | Consider a space object in an orbit about the earth. An uncertain initial
state can be represented as a point cloud which can be propagated to later
times by the laws of Newtonian motion. If the state of the object is
represented in Cartesian earth centered inertial (Cartesian-ECI) coordinates,
then even if initial uncertainty is Gaussian in this coordinate system, the
distribution quickly becomes non-Gaussian as the propagation time increases.
Similar problems arise in other standard fixed coordinate systems in
astrodynamics, e.g. Keplerian and to some extent equinoctial. To address these
problems, a local "Adapted STructural (AST)'' coordinate system has been
developed in which uncertainty is represented in terms of deviations from a
"central state".
Given a sequence of angles-only measurements, the iterated nonlinear extended
(IEKF) and unscented (IUKF) Kalman filters are often the most appropriate
variants to use. In particular, they can be much more accurate than the more
commonly used non-iterated versions, the extended (EKF) and unscented (UKF)
Kalman filters, especially under high eccentricity. In addition, iterated
Kalman filters can often be well-approximated by two new closed form filters,
the observation-centered extended (OCEKF) and unscented (OCUKF) Kalman filters.
| stat |
Efficient and minimal length parametric conformal prediction regions | Conformal prediction methods construct prediction regions for iid data that
are valid in finite samples. We provide two parametric conformal prediction
regions that are applicable for a wide class of continuous statistical models.
This class of statistical models includes generalized linear models (GLMs) with
continuous outcomes. Our parametric conformal prediction regions possesses
finite sample validity, even when the model is misspecified, and are
asymptotically of minimal length when the model is correctly specified. The
first parametric conformal prediction region is constructed through binning of
the predictor space, guarantees finite-sample local validity and is
asymptotically minimal at the $\sqrt{\log(n)/n}$ rate when the dimension $d$ of
the predictor space is one or two, and converges at the
$O\{(\log(n)/n)^{1/d}\}$ rate when $d > 2$. The second parametric conformal
prediction region is constructed by transforming the outcome variable to a
common distribution via the probability integral transform, guarantees
finite-sample marginal validity, and is asymptotically minimal at the
$\sqrt{\log(n)/n}$ rate. We develop a novel concentration inequality for
maximum likelihood estimation that induces these convergence rates. We analyze
prediction region coverage properties, large-sample efficiency, and robustness
properties of four methods for constructing conformal prediction intervals for
GLMs: fully nonparametric kernel-based conformal, residual based conformal,
normalized residual based conformal, and parametric conformal which uses the
assumed GLM density as a conformity measure. Extensive simulations compare
these approaches to standard asymptotic prediction regions. The utility of the
parametric conformal prediction region is demonstrated in an application to
interval prediction of glycosylated hemoglobin levels, a blood measurement used
to diagnose diabetes.
| stat |
Extreme Compressive Sampling for Covariance Estimation | This paper studies the problem of estimating the covariance of a collection
of vectors using only highly compressed measurements of each vector. An
estimator based on back-projections of these compressive samples is proposed
and analyzed. A distribution-free analysis shows that by observing just a
single linear measurement of each vector, one can consistently estimate the
covariance matrix, in both infinity and spectral norm, and this same analysis
leads to precise rates of convergence in both norms. Via information-theoretic
techniques, lower bounds showing that this estimator is minimax-optimal for
both infinity and spectral norm estimation problems are established. These
results are also specialized to give matching upper and lower bounds for
estimating the population covariance of a collection of Gaussian vectors, again
in the compressive measurement model. The analysis conducted in this paper
shows that the effective sample complexity for this problem is scaled by a
factor of $m^2/d^2$ where $m$ is the compression dimension and $d$ is the
ambient dimension. Applications to subspace learning (Principal Components
Analysis) and learning over distributed sensor networks are also discussed.
| stat |
Context Vectors are Reflections of Word Vectors in Half the Dimensions | This paper takes a step towards theoretical analysis of the relationship
between word embeddings and context embeddings in models such as word2vec. We
start from basic probabilistic assumptions on the nature of word vectors,
context vectors, and text generation. These assumptions are well supported
either empirically or theoretically by the existing literature. Next, we show
that under these assumptions the widely-used word-word PMI matrix is
approximately a random symmetric Gaussian ensemble. This, in turn, implies that
context vectors are reflections of word vectors in approximately half the
dimensions. As a direct application of our result, we suggest a theoretically
grounded way of tying weights in the SGNS model.
| stat |
Statistically Guided Divide-and-Conquer for Sparse Factorization of
Large Matrix | The sparse factorization of a large matrix is fundamental in modern
statistical learning. In particular, the sparse singular value decomposition
and its variants have been utilized in multivariate regression, factor
analysis, biclustering, vector time series modeling, among others. The appeal
of this factorization is owing to its power in discovering a
highly-interpretable latent association network, either between samples and
variables or between responses and predictors. However, many existing methods
are either ad hoc without a general performance guarantee, or are
computationally intensive, rendering them unsuitable for large-scale studies.
We formulate the statistical problem as a sparse factor regression and tackle
it with a divide-and-conquer approach. In the first stage of division, we
consider both sequential and parallel approaches for simplifying the task into
a set of co-sparse unit-rank estimation (CURE) problems, and establish the
statistical underpinnings of these commonly-adopted and yet poorly understood
deflation methods. In the second stage of division, we innovate a contended
stagewise learning technique, consisting of a sequence of simple incremental
updates, to efficiently trace out the whole solution paths of CURE. Our
algorithm has a much lower computational complexity than alternating convex
search, and the choice of the step size enables a flexible and principled
tradeoff between statistical accuracy and computational efficiency. Our work is
among the first to enable stagewise learning for non-convex problems, and the
idea can be applicable in many multi-convex problems. Extensive simulation
studies and an application in genetics demonstrate the effectiveness and
scalability of our approach.
| stat |
Dealing with Nuisance Parameters using Machine Learning in High Energy
Physics: a Review | In this work we discuss the impact of nuisance parameters on the
effectiveness of machine learning in high-energy physics problems, and provide
a review of techniques that allow to include their effect and reduce their
impact in the search for optimal selection criteria and variable
transformations. The introduction of nuisance parameters complicates the
supervised learning task and its correspondence with the data analysis goal,
due to their contribution degrading the model performances in real data, and
the necessary addition of uncertainties in the resulting statistical inference.
The approaches discussed include nuisance-parameterized models, modified or
adversary losses, semi-supervised learning approaches, and inference-aware
techniques.
| stat |
Smooth Contextual Bandits: Bridging the Parametric and
Non-differentiable Regret Regimes | We study a nonparametric contextual bandit problem where the expected reward
functions belong to a H\"older class with smoothness parameter $\beta$. We show
how this interpolates between two extremes that were previously studied in
isolation: non-differentiable bandits ($\beta\leq1$), where rate-optimal regret
is achieved by running separate non-contextual bandits in different context
regions, and parametric-response bandits (satisfying $\beta=\infty$), where
rate-optimal regret can be achieved with minimal or no exploration due to
infinite extrapolatability. We develop a novel algorithm that carefully adjusts
to all smoothness settings and we prove its regret is rate-optimal by
establishing matching upper and lower bounds, recovering the existing results
at the two extremes. In this sense, our work bridges the gap between the
existing literature on parametric and non-differentiable contextual bandit
problems and between bandit algorithms that exclusively use global or local
information, shedding light on the crucial interplay of complexity and regret
in contextual bandits.
| stat |
Inferring Risks of Coronavirus Transmission from Community Household
Data | The response of many governments to the COVID-19 pandemic has involved
measures to control within- and between-household transmission, providing
motivation to improve understanding of the absolute and relative risks in these
contexts. Here, we perform exploratory, residual-based, and
transmission-dynamic household analysis of the Office for National Statistics
(ONS) COVID-19 Infection Survey (CIS) data from 26 April 2020 to 8 March 2021
in England. This provides evidence for: (i) temporally varying rates of
introduction of infection into households broadly following the trajectory of
the overall epidemic; (ii) Susceptible-Infectious Transmission Probabilities
(SITPs) of within-household transmission in the 15-35% range; (iii) the
emergence of the B.1.1.7 variant, being around 50% more infectious within
households; (iv) significantly (in the range 25-300%) more risk of bringing
infection into the household for workers in patient-facing roles; (v) increased
risk for secondary school-age children of bringing the infection into the
household when schools are open (in the range 64-235%); (vi) increased risk for
primary school-age children of bringing the infection into the household when
schools were open in late autumn 2020 (around 40%).
| stat |
Common Statistical Patterns in Urban Terrorism | The underlying reasons behind modern terrorism are seemingly complex and
intangible. Despite diverse causal mechanisms, research has shown that there
exists general statistical patterns at the global scale that can shed light on
human confrontation behaviour. Whilst many policing and counter-terrorism
operations are conducted at a city level, there has been a lack of research in
building city-level resolution prediction engines based on statistical
patterns. For the first time, the paper shows that there exists general
commonalities between global cities under terrorist attacks. By examining over
30,000 geo-tagged terrorism acts over 7000 cities worldwide from 2002 to today,
the results shows the following. All cities experience attacks $A$ that are
uncorrelated to the population and separated by a time interval $t$ that is
negative exponentially distributed $\sim \exp(-A^{-1})$, with a death-toll per
attack that follows a power law distribution. The prediction parameters yield a
high confidence of explaining up to 87\% of the variations in frequency and
89\% in the death-toll data. These findings show that the aggregate statistical
behaviour of terror attacks are seemingly random and memoryless for all global
cities. The enabled the author to develop a data-driven city-specific
prediction system and we quantify its information theoretic uncertainty and
information loss. Further analysis show that there appears to be an increase in
the uncertainty over the predictability of attacks, challenging our ability to
develop effective counter-terrorism capabilities.
| stat |
Use of Computer Applications for Determining the Best Possible Runway
Orientation using Wind Rose Diagrams | As technology advances, there are more uses of computer applications in a
Civil Engineering Projects. To simplify the work, minimize errors, and obviate
computational time required in designing and other structural analysis work,
software which give accurate results to the given inputs are the need of the
hour and this paper highlights the simulation of a manual procedure to find out
the orientation of a Runway from the Wind Data available. A Computer
application in Java on Net beans platform has been used to develop this Wind
Rose Software. This software has been developed using the concepts of Wind Rose
Diagram Type-II which will calculate the best possible orientation of the
Runway along with the Coverage.
| stat |
Unbounded Slice Sampling | Slice sampling is an efficient Markov Chain Monte Carlo algorithm to sample
from an unnormalized density with acceptance ratio always $1$. However, when
the variable to sample is unbounded, its "stepping-out" heuristic works only
locally, making it difficult to uniformly explore possible candidates. This
paper proposes a simple change-of-variable method to slice sample an unbounded
variable equivalently from [0,1).
| stat |
Estimating network memberships by simplex vertex hunting | Consider an undirected mixed membership network with $n$ nodes and $K$
communities. For each node $i$, $1 \leq i \leq n$, we model the membership by a
Probability Mass Function (PMF) $\pi_{i} = (\pi_{i}(1), \pi_{i}(2), \ldots$,
$\pi_{i}(K))'$, where $\pi_{i}(k)$ is the probability that node $i$ belongs to
community $k$, $1 \leq k \leq K$. We call node $i$ "pure" if $\pi_i$ is
degenerate and "mixed" otherwise. The primary interest is to estimate $\pi_i$,
$1 \leq i \leq n$.
We model the adjacency matrix $A$ with a Degree Corrected Mixed Membership
(DCMM) model, and let $\hat{\xi}_1, \hat{\xi}_2, \ldots, \hat{\xi}_K$ be the
first $K$ eigenvectors. Define a matrix $\hat{R} \in \mathbb{R}^{n, K-1}$ by
$\hat{R}(i,k) = \hat{\xi}_{k+1}(i)/\hat{\xi}_1(i)$, $1 \leq k \leq K-1$, $1
\leq i \leq n$. The oracle counterpart of $\hat{R}$ (denoted by $R$) under the
DCMM model contains all information we need for the memberships. In fact, we
have an interesting insight: there is a simplex ${\cal S}$ in
$\mathbb{R}^{K-1}$ such that row $i$ of $R$ corresponds to a vertex of ${\cal
S}$ if node $i$ is pure, and corresponds to an interior point of ${\cal S}$
otherwise. Vertex Hunting (i.e., estimating the vertices of ${\cal S}$) is
therefore the key to our problem.
We propose a new approach Mixed-SCORE to membership estimation, with an
easy-to-use Vertex Hunting step. We derive the convergence rate of Mixed-SCORE
using delicate spectral analysis, especially tight row-wise deviation bounds
for $\hat{R}$. We have also applied it to $4$ network data sets with
encouraging results.
| stat |
Bridging Convex and Nonconvex Optimization in Robust PCA: Noise,
Outliers, and Missing Data | This paper delivers improved theoretical guarantees for the convex
programming approach in low-rank matrix estimation, in the presence of (1)
random noise, (2) gross sparse outliers, and (3) missing data. This problem,
often dubbed as robust principal component analysis (robust PCA), finds
applications in various domains. Despite the wide applicability of convex
relaxation, the available statistical support (particularly the stability
analysis vis-\`a-vis random noise) remains highly suboptimal, which we
strengthen in this paper. When the unknown matrix is well-conditioned,
incoherent, and of constant rank, we demonstrate that a principled convex
program achieves near-optimal statistical accuracy, in terms of both the
Euclidean loss and the $\ell_{\infty}$ loss. All of this happens even when
nearly a constant fraction of observations are corrupted by outliers with
arbitrary magnitudes. The key analysis idea lies in bridging the convex program
in use and an auxiliary nonconvex optimization algorithm, and hence the title
of this paper.
| stat |
Inference and Computation for Sparsely Sampled Random Surfaces | Non-parametric inference for functional data over two-dimensional domains
entails additional computational and statistical challenges, compared to the
one-dimensional case. Separability of the covariance is commonly assumed to
address these issues in the densely observed regime. Instead, we consider the
sparse regime, where the latent surfaces are observed only at few irregular
locations with additive measurement error, and propose an estimator of
covariance based on local linear smoothers. Consequently, the assumption of
separability reduces the intrinsically four-dimensional smoothing problem into
several two-dimensional smoothers and allows the proposed estimator to retain
the classical minimax-optimal convergence rate for two-dimensional smoothers.
Even when separability fails to hold, imposing it can be still advantageous as
a form of regularization. A simulation study reveals a favorable bias-variance
trade-off and massive speed-ups achieved by our approach. Finally, the proposed
methodology is used for qualitative analysis of implied volatility surfaces
corresponding to call options, and for prediction of the latent surfaces based
on information from the entire data set, allowing for uncertainty
quantification. Our cross-validated out-of-sample quantitative results show
that the proposed methodology outperforms the common approach of pre-smoothing
every implied volatility surface separately.
| stat |
Governance on Social Media Data: Different Focuses between Government
and Internet Company | How governments and Internet companies regulate user data on social media
attracts public attention. This study tried to answer two questions: What kind
of countries send more requests for Facebook user data? What kind of countries
get more requests replies from Facebook? We aim to figure out how a country's
economic, political and social factors affect its government requests for user
data and Facebook's responses rate to those requests. Results show that
countries with higher GDP per capita, a higher level of human freedom and a
lower level of rule of law send more requests for user data; while Facebook
tends to reply to government requests from countries with a higher level of
human freedom and a lower level of political stability. In conclusion,
governments and Facebook show different focuses on governance on social media
data.
| stat |
Computationally Efficient Estimation of the Spectral Gap of a Markov
Chain | We consider the problem of estimating from sample paths the absolute spectral
gap $\gamma_*$ of a reversible, irreducible and aperiodic Markov chain
$(X_t)_{t \in \mathbb{N}}$ over a finite state space $\Omega$. We propose the
${\tt UCPI}$ (Upper Confidence Power Iteration) algorithm for this problem, a
low-complexity algorithm which estimates the spectral gap in time ${\cal O}(n)$
and memory space ${\cal O}((\ln n)^2)$ given $n$ samples. This is in stark
contrast with most known methods which require at least memory space ${\cal
O}(|\Omega|)$, so that they cannot be applied to large state spaces.
Furthermore, ${\tt UCPI}$ is amenable to parallel implementation.
| stat |
Randomized and Balanced Allocation of Units into Treatment Groups Using
the Finite Selection Model for R | The original Finite Selection Model (FSM) was developed in the 1970s to
enhance the design of the RAND Health Insurance Experiment (HIE; Newhouse et
al. 1993). At the time of its development by Carl Morris (Morris 1979), there
were fundamental computational limitations to make the method widely available
for practitioners. Today, as randomized experiments increasingly become more
common, there is a need for implementing experimental designs that are
randomized, balanced, robust, and easily applicable to several treatment
groups. To help address this problem, we revisit the original FSM under the
potential outcome framework for causal inference and provide its first readily
available software implementation. In this paper, we provide an introduction to
the FSM and a step-by-step guide for its use in R.
| stat |
Efficient Inference of Flexible Interaction in Spiking-neuron Networks | Hawkes process provides an effective statistical framework for analyzing the
time-dependent interaction of neuronal spiking activities. Although utilized in
many real applications, the classic Hawkes process is incapable of modelling
inhibitory interactions among neurons. Instead, the nonlinear Hawkes process
allows for a more flexible influence pattern with excitatory or inhibitory
interactions. In this paper, three sets of auxiliary latent variables
(P\'{o}lya-Gamma variables, latent marked Poisson processes and sparsity
variables) are augmented to make functional connection weights in a Gaussian
form, which allows for a simple iterative algorithm with analytical updates. As
a result, an efficient expectation-maximization (EM) algorithm is derived to
obtain the maximum a posteriori (MAP) estimate. We demonstrate the accuracy and
efficiency performance of our algorithm on synthetic and real data. For real
neural recordings, we show our algorithm can estimate the temporal dynamics of
interaction and reveal the interpretable functional connectivity underlying
neural spike trains.
| stat |
Nearest Neighbor Search Under Uncertainty | Nearest Neighbor Search (NNS) is a central task in knowledge representation,
learning, and reasoning. There is vast literature on efficient algorithms for
constructing data structures and performing exact and approximate NNS. This
paper studies NNS under Uncertainty (NNSU). Specifically, consider the setting
in which an NNS algorithm has access only to a stochastic distance oracle that
provides a noisy, unbiased estimate of the distance between any pair of points,
rather than the exact distance. This models many situations of practical
importance, including NNS based on human similarity judgements, physical
measurements, or fast, randomized approximations to exact distances. A naive
approach to NNSU could employ any standard NNS algorithm and repeatedly query
and average results from the stochastic oracle (to reduce noise) whenever it
needs a pairwise distance. The problem is that a sufficient number of repeated
queries is unknown in advance; e.g., a point maybe distant from all but one
other point (crude distance estimates suffice) or it may be close to a large
number of other points (accurate estimates are necessary). This paper shows how
ideas from cover trees and multi-armed bandits can be leveraged to develop an
NNSU algorithm that has optimal dependence on the dataset size and the
(unknown)geometry of the dataset.
| stat |
Dating the Break in High-dimensional Data | This paper is concerned with estimation and inference for the location of a
change point in the mean of independent high-dimensional data. Our change point
location estimator maximizes a new U-statistic based objective function, and
its convergence rate and asymptotic distribution after suitable centering and
normalization are obtained under mild assumptions. Our estimator turns out to
have better efficiency as compared to the least squares based counterpart in
the literature. Based on the asymptotic theory, we construct a confidence
interval by plugging in consistent estimates of several quantities in the
normalization. We also provide a bootstrap-based confidence interval and state
its asymptotic validity under suitable conditions. Through simulation studies,
we demonstrate favorable finite sample performance of the new change point
location estimator as compared to its least squares based counterpart, and our
bootstrap-based confidence intervals, as compared to several existing
competitors. The asymptotic theory based on high-dimensional U-statistic is
substantially different from those developed in the literature and is of
independent interest.
| stat |
Lasso, knockoff and Gaussian covariates: a comparison | Given data $\mathbf{y}$ and $k$ covariates $\mathbf{x}_j$ one problem in
linear regression is to decide which if any of the covariates to include when
regressing the dependent variable $\mathbf{y}$ on the covariates
$\mathbf{x}_j$. In this paper three such methods, lasso, knockoff and Gaussian
covariates are compared using simulations and real data. The Gaussian covariate
method is based on exact probabilities which are valid for all $\mathbf{y}$ and
$\mathbf{x}_j$ making it model free. Moreover the probabilities agree with
those based on the F-distribution for the standard linear model with i.i.d.
Gaussian errors. It is conceptually, mathematically and algorithmically very
simple, it is very fast and makes no use of simulations. It outperforms lasso
and knockoff in all respects by a considerable margin.
| stat |
On greedy heuristics for computing D-efficient saturated subsets | Let $\mathcal{F}$ be a set consisting of $n$ real vectors of dimension $m
\leq n$. For any saturated, i.e., $m$-element, subset $\mathcal{S}$ of
$\mathcal{F}$, let $\mathrm{vol}(\mathcal{S})$ be the volume of the
parallelotope formed by the vectors of $\mathcal{S}$. A set $\mathcal{S}^*$ is
called a $D$-optimal saturated subset of $\mathcal{F}$, if it maximizes
$\mathrm{vol}(\mathcal{S})$ among all saturated subsets of $\mathcal{F}$. In
this paper, we propose two greedy heuristics for the construction of saturated
subsets performing well with respect to the criterion of $D$-optimality: an
improvement of the method suggested by Galil and Kiefer for the initiation of
$D$-optimal experimental design algorithms, and a modification of the
Kumar-Yildirim method, the original version of which was proposed for the
initiation of the minimum-volume enclosing ellipsoid algorithms. We provide
geometric and analytic insights into the two methods, and compare them to the
commonly used random and regularized greedy heuristics. We also suggest
variants of the greedy methods for a large set $\mathcal{F}$, for the
construction of $D$-efficient non-saturated subsets, and for alternative
optimality criteria.
| stat |
Saddlepoint-adjusted inversion of characteristic functions | For certain types of statistical models, the characteristic function (Fourier
transform) is available in closed form, whereas the probability density
function has an intractable form, typically as an infinite sum of probability
weighted densities. Important such examples include solutions of stochastic
differential equations with jumps, the Tweedie model, and Poisson mixture
models. We propose a novel and general numerical method for retrieving the
probability density function from the characteristic function, conditioned on
the existence of the moment generating function. Unlike methods based on direct
application of quadrature to the inverse Fourier transform, the proposed method
allows accurate evaluation of the log-probability density function arbitrarily
far out in the tail. Moreover, unlike ordinary saddlepoint approximations, the
proposed methodology is in principle exact modulus discretization and
truncation error of quadrature applied to inversion in a high-density region.
Owing to these properties, the proposed method is computationally stable and
very accurate under log-likelihood optimisation. The method is illustrated for
a normal variance-mean mixture, and in an application of maximum likelihood
estimation to a jump diffusion model for financial data.
| stat |
A Robust Interrupted Time Series Model for Analyzing Complex Healthcare
Intervention Data | Current health policy calls for greater use of evidence based care delivery
services to improve patient quality and safety outcomes. Care delivery is
complex, with interacting and interdependent components that challenge
traditional statistical analytic techniques, in particular when modeling a time
series of outcomes data that might be "interrupted" by a change in a particular
method of health care delivery. Interrupted time series (ITS) is a robust
quasi-experimental design with the ability to infer the effectiveness of an
intervention that accounts for data dependency. Current standardized methods
for analyzing ITS data do not model changes in variation and correlation
following the intervention. This is a key limitation since it is plausible for
data variability and dependency to change because of the intervention.
Moreover, present methodology either assumes a pre-specified interruption time
point with an instantaneous effect or removes data for which the effect of
intervention is not fully realized. In this paper, we describe and develop a
novel `Robust-ITS' model that overcomes these omissions and limitations. The
Robust-ITS model formally performs inference on: (a) identifying the change
point; (b) differences in pre- and post-intervention correlation; (c)
differences in the outcome variance pre- and post-intervention; and (d)
differences in the mean pre- and post-intervention. We illustrate the proposed
method by analyzing patient satisfaction data from a hospital that implemented
and evaluated a new nursing care delivery model as the intervention of
interest. The Robust-ITS model is implemented in a R Shiny toolbox which is
freely available to the community.
| stat |
Understanding Double Descent Requires a Fine-Grained Bias-Variance
Decomposition | Classical learning theory suggests that the optimal generalization
performance of a machine learning model should occur at an intermediate model
complexity, with simpler models exhibiting high bias and more complex models
exhibiting high variance of the predictive function. However, such a simple
trade-off does not adequately describe deep learning models that simultaneously
attain low bias and variance in the heavily overparameterized regime. A primary
obstacle in explaining this behavior is that deep learning algorithms typically
involve multiple sources of randomness whose individual contributions are not
visible in the total variance. To enable fine-grained analysis, we describe an
interpretable, symmetric decomposition of the variance into terms associated
with the randomness from sampling, initialization, and the labels. Moreover, we
compute the high-dimensional asymptotic behavior of this decomposition for
random feature kernel regression, and analyze the strikingly rich phenomenology
that arises. We find that the bias decreases monotonically with the network
width, but the variance terms exhibit non-monotonic behavior and can diverge at
the interpolation boundary, even in the absence of label noise. The divergence
is caused by the \emph{interaction} between sampling and initialization and can
therefore be eliminated by marginalizing over samples (i.e. bagging) \emph{or}
over the initial parameters (i.e. ensemble learning).
| stat |
Robustly Extracting Medical Knowledge from EHRs: A Case Study of
Learning a Health Knowledge Graph | Increasingly large electronic health records (EHRs) provide an opportunity to
algorithmically learn medical knowledge. In one prominent example, a causal
health knowledge graph could learn relationships between diseases and symptoms
and then serve as a diagnostic tool to be refined with additional clinical
input. Prior research has demonstrated the ability to construct such a graph
from over 270,000 emergency department patient visits. In this work, we
describe methods to evaluate a health knowledge graph for robustness. Moving
beyond precision and recall, we analyze for which diseases and for which
patients the graph is most accurate. We identify sample size and unmeasured
confounders as major sources of error in the health knowledge graph. We
introduce a method to leverage non-linear functions in building the causal
graph to better understand existing model assumptions. Finally, to assess model
generalizability, we extend to a larger set of complete patient visits within a
hospital system. We conclude with a discussion on how to robustly extract
medical knowledge from EHRs.
| stat |
Continuous normalizing flows on manifolds | Normalizing flows are a powerful technique for obtaining reparameterizable
samples from complex multimodal distributions. Unfortunately, current
approaches are only available for the most basic geometries and fall short when
the underlying space has a nontrivial topology, limiting their applicability
for most real-world data. Using fundamental ideas from differential geometry
and geometric control theory, we describe how the recently introduced Neural
ODEs and continuous normalizing flows can be extended to arbitrary smooth
manifolds. We propose a general methodology for parameterizing vector fields on
these spaces and demonstrate how gradient-based learning can be performed.
Additionally, we provide a scalable unbiased estimator for the divergence in
this generalized setting. Experiments on a diverse selection of spaces
empirically showcase the defined framework's ability to obtain
reparameterizable samples from complex distributions.
| stat |
A Machine Learning Approach to Shipping Box Design | Having the right assortment of shipping boxes in the fulfillment warehouse to
pack and ship customer's online orders is an indispensable and integral part of
nowadays eCommerce business, as it will not only help maintain a profitable
business but also create great experiences for customers. However, it is an
extremely challenging operations task to strategically select the best
combination of tens of box sizes from thousands of feasible ones to be
responsible for hundreds of thousands of orders daily placed on millions of
inventory products. In this paper, we present a machine learning approach to
tackle the task by formulating the box design problem prescriptively as a
generalized version of weighted $k$-medoids clustering problem, where the
parameters are estimated through a variety of descriptive analytics. We test
this machine learning approach on fulfillment data collected from Walmart U.S.
eCommerce, and our approach is shown to be capable of improving the box
utilization rate by more than $10\%$.
| stat |
Bayesian Inference for Polya Inverse Gamma Models | Probability density functions that include the gamma function are widely used
in statistics and machine learning. The normalizing constants of gamma, inverse
gamma, beta, and Dirichlet distributions all include model parameters as
arguments in the gamma function; however, the gamma function does not naturally
admit a conjugate prior distribution in a Bayesian analysis, and statistical
inference of these parameters is a significant challenge. In this paper, we
construct the Polya-inverse Gamma (P-IG) distribution as an infinite
convolution of Generalized inverse Gaussian (GIG) distributions, and we
represent the reciprocal gamma function as a scale mixture of normal
distributions. As a result, the P-IG distribution yields an efficient data
augmentation strategy for fully Bayesian inference on model parameters in
gamma, inverse gamma, beta, and Dirichlet distributions. To illustrate the
applied utility of our data augmentation strategy, we infer the proportion of
overdose deaths in the United States attributed to different opioid and
prescription drugs with a Dirichlet allocation model.
| stat |
CO-Optimal Transport | Optimal transport (OT) is a powerful geometric and probabilistic tool for
finding correspondences and measuring similarity between two distributions.
Yet, its original formulation relies on the existence of a cost function
between the samples of the two distributions, which makes it impractical when
they are supported on different spaces. To circumvent this limitation, we
propose a novel OT problem, named COOT for CO-Optimal Transport, that
simultaneously optimizes two transport maps between both samples and features,
contrary to other approaches that either discard the individual features by
focusing on pairwise distances between samples or need to model explicitly the
relations between them. We provide a thorough theoretical analysis of our
problem, establish its rich connections with other OT-based distances and
demonstrate its versatility with two machine learning applications in
heterogeneous domain adaptation and co-clustering/data summarization, where
COOT leads to performance improvements over the state-of-the-art methods.
| stat |
Smoothing splines on Riemannian manifolds, with applications to 3D shape
space | There has been increasing interest in statistical analysis of data lying in
manifolds. This paper generalizes a smoothing spline fitting method to
Riemannian manifold data based on the technique of unrolling and unwrapping
originally proposed in Jupp and Kent (1987) for spherical data. In particular
we develop such a fitting procedure for shapes of configurations in general
$m$-dimensional Euclidean space, extending our previous work for two
dimensional shapes. We show that parallel transport along a geodesic on Kendall
shape space is linked to the solution of a homogeneous first-order differential
equation, some of whose coefficients are implicitly defined functions. This
finding enables us to approximate the procedure of unrolling and unwrapping by
simultaneously solving such equations numerically, and so to find numerical
solutions for smoothing splines fitted to higher dimensional shape data. This
fitting method is applied to the analysis of some dynamic 3D peptide data.
| stat |
Estimating the radii of air bubbles in water using passive acoustic
monitoring | The study of the acoustic emission of underwater gas bubbles is a subject of
both theoretical and applied interest, since it finds an important application
in the development of acoustic monitoring tools for detection and
quantification of underwater gas leakages. An underlying physical model is
essential in the study of such emissions, but is not enough: also some
statistical procedure must be applied in order to deal with all uncertainties
(including those caused by background noise). In this paper we take a
probabilistic (Bayesian) methodology which is well known in the statistical
signal analysis communitiy, and apply it to the problem of estimating the radii
of air bubbles in water. We introduce the bubblegram, a feature extraction
technique graphically similar to the traditional spectrogram but tailored to
respond only to pulse structures that correspond to a given physical model. We
investigate the performance of the bubblegram and our model in general using
laboratory generated data.
| stat |
Eigenvalue and Generalized Eigenvalue Problems: Tutorial | This paper is a tutorial for eigenvalue and generalized eigenvalue problems.
We first introduce eigenvalue problem, eigen-decomposition (spectral
decomposition), and generalized eigenvalue problem. Then, we mention the
optimization problems which yield to the eigenvalue and generalized eigenvalue
problems. We also provide examples from machine learning, including principal
component analysis, kernel supervised principal component analysis, and Fisher
discriminant analysis, which result in eigenvalue and generalized eigenvalue
problems. Finally, we introduce the solutions to both eigenvalue and
generalized eigenvalue problems.
| stat |
On the cross-validation bias due to unsupervised pre-processing | Cross-validation is the de facto standard for predictive model evaluation and
selection. In proper use, it provides an unbiased estimate of a model's
predictive performance. However, data sets often undergo various forms of
data-dependent preprocessing, such as mean-centering, rescaling, dimensionality
reduction, and outlier removal. It is often believed that such preprocessing
stages, if done in an unsupervised manner (that does not incorporate the class
labels or response values) are generally safe to do prior to cross-validation.
In this paper, we study three commonly-practiced preprocessing procedures prior
to a regression analysis: (i) variance-based feature selection; (ii) grouping
of rare categorical features; and (iii) feature rescaling. We demonstrate that
unsupervised preprocessing procedures can, in fact, introduce a large bias into
cross-validation estimates and potentially lead to sub-optimal model selection.
This bias may be either positive or negative and its exact magnitude depends on
all the parameters of the problem in an intricate manner. Further research is
needed to understand the real-world impact of this bias across different
application domains, particularly when dealing with low sample counts and
high-dimensional data.
| stat |
Causal version of Principle of Insufficient Reason and MaxEnt | The Principle of insufficient Reason (PIR) assigns equal probabilities to
each alternative of a random experiment whenever there is no reason to prefer
one over the other. Maximum Entropy (MaxEnt) generalizes PIR to the case where
statistical information like expectations are given. It is known that both
principles result in paradox probability updates for joint distributions of
cause and effect. This is because constraints on the conditional P(effect |
cause) result in changes of P(cause) that assign higher probability to those
values of the cause that offer more options for the effect, suggesting
'intentional behaviour'. Earlier work therefore suggested sequentially
maximizing (conditional) entropy according to the causal order, but without
further justification apart from plausibility for toy examples. We justify
causal modifications of PIR and MaxEnt by separating constraints into
restrictions for the cause and restrictions for the mechanism that generates
the effect from the cause. We further sketch why Causal PIR also entails
'Information Geometric Causal Inference'.
We briefly discuss problems of generalizing the causal version of MaxEnt to
arbitrary causal DAGs.
| stat |
Sparse Tensor Additive Regression | Tensors are becoming prevalent in modern applications such as medical imaging
and digital marketing. In this paper, we propose a sparse tensor additive
regression (STAR) that models a scalar response as a flexible nonparametric
function of tensor covariates. The proposed model effectively exploits the
sparse and low-rank structures in the tensor additive regression. We formulate
the parameter estimation as a non-convex optimization problem, and propose an
efficient penalized alternating minimization algorithm. We establish a
non-asymptotic error bound for the estimator obtained from each iteration of
the proposed algorithm, which reveals an interplay between the optimization
error and the statistical rate of convergence. We demonstrate the efficacy of
STAR through extensive comparative simulation studies, and an application to
the click-through-rate prediction in online advertising.
| stat |
Prediction of Spatial Point Processes: Regularized Method with
Out-of-Sample Guarantees | A spatial point process can be characterized by an intensity function which
predicts the number of events that occur across space. In this paper, we
develop a method to infer predictive intensity intervals by learning a spatial
model using a regularized criterion. We prove that the proposed method exhibits
out-of-sample prediction performance guarantees which, unlike standard
estimators, are valid even when the spatial model is misspecified. The method
is demonstrated using synthetic as well as real spatial data.
| stat |
Single-Model Uncertainties for Deep Learning | We provide single-model estimates of aleatoric and epistemic uncertainty for
deep neural networks. To estimate aleatoric uncertainty, we propose
Simultaneous Quantile Regression (SQR), a loss function to learn all the
conditional quantiles of a given target variable. These quantiles can be used
to compute well-calibrated prediction intervals. To estimate epistemic
uncertainty, we propose Orthonormal Certificates (OCs), a collection of diverse
non-constant functions that map all training samples to zero. These
certificates map out-of-distribution examples to non-zero values, signaling
epistemic uncertainty. Our uncertainty estimators are computationally
attractive, as they do not require ensembling or retraining deep models, and
achieve competitive performance.
| stat |
Interval-Valued Kriging Models for Geostatistical Mapping with Imprecise
Inputs | Many geosciences data are imprecise due to various limitations and
uncertainties in the measuring process. One way to preserve this imprecision in
a geostatistical mapping framework is to characterize the measurements as
intervals rather than single values. To effectively analyze the interval-valued
data, this paper proposes and develops interval-valued kriging models based on
the theory of random sets and a generalized L2 metric. These models overcome
the mathematical difficulties of a previous development and are computationally
more feasible. Numerical implementation of our interval-valued kriging is
provided using a penalty-based constrained optimization algorithm. An
application to the prediction of design ground snow loads in Utah, USA, is
presented that demonstrates the advantages of the proposed models in preserving
crucial sources of uncertainty towards a more efficient risk-based designing
scheme.
| stat |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.