Search arXiv⌕ Search

arXiv subjects

David Draper

Publications and source records attributed to David Draper.

At least 19 recordsLinked to original sources
arXiv2024

A Unified Framework for Cluster Methods with Tensor Networks

Markov Chain Monte Carlo (MCMC), and Tensor Networks (TN) are two powerful frameworks for numerically investigating many-body systems, each offering distinct advantages. MCMC, with its flexibility and theoretical consistency, is well-suited for simulating arbitrary systems by sampling. TN, on the other hand, provides a powerful tensor-based language for capturing the entanglement properties intrinsic to many-body systems, offering a universal representation of these systems. In this work, we leverage the computational strengths of TN to design a versatile cluster MCMC sampler. Specifically, we propose a general framework for constructing tensor-based cluster MCMC methods, enabling arbitrary cluster updates by utilizing TNs to compute the distributions required in the MCMC sampler. Our framework unifies several existing cluster algorithms as special cases and allows for natural extensions. We demonstrate our method by applying it to the simulation of the two-dimensional Edwards-Anderson Model and the three-dimensional Ising Model. This work is dedicated to the memory of Prof. David Draper.

stat.ME↗
arXiv2022

Annealing Double-Head: An Architecture for Online Calibration of Deep Neural Networks

Model calibration, which is concerned with how frequently the model predicts correctly, not only plays a vital part in statistical model design, but also has substantial practical applications, such as optimal decision-making in the real world. However, it has been discovered that modern deep neural networks are generally poorly calibrated due to the overestimation (or underestimation) of predictive confidence, which is closely related to overfitting. In this paper, we propose Annealing Double-Head, a simple-to-implement but highly effective architecture for calibrating the DNN during training. To be precise, we construct an additional calibration head-a shallow neural network that typically has one latent layer-on top of the last latent layer in the normal model to map the logits to the aligned confidence. Furthermore, a simple Annealing technique that dynamically scales the logits by calibration head in training procedure is developed to improve its performance. Under both the in-distribution and distributional shift circumstances, we exhaustively evaluate our Annealing Double-Head architecture on multiple pairs of contemporary DNN architectures and vision and speech datasets. We demonstrate that our method achieves state-of-the-art model calibration performance without post-processing while simultaneously providing comparable predictive accuracy in comparison to other recently proposed calibration methods on a range of learning tasks.

stat.ML↗
arXiv2021

A Simple Necessary Condition For Independence of Real-Valued Random Variables

The standard method to check for the independence of two real-valued random variables -- demonstrating that the bivariate joint distribution factors into the product of its marginals -- is both necessary and sufficient. Here we present a simple necessary condition based on the support sets of the random variables, which -- if not satisfied -- avoids the need to extract the marginals from the joint in demonstrating dependence. We review, in an accessible manner, the measure-theoretic, topological, and probabilistic details necessary to establish the background for the old and new ideas presented here. We prove our result in both the discrete case (where the basic ideas emerge in a simple setting), the continuous case (where serious complications emerge), and for general real-valued random variables, and we illustrate the use of our condition in three simple examples.

math.PR↗
arXiv2021

Neural Tangent Kernel of Matrix Product States: Convergence and Applications

In this work, we study the Neural Tangent Kernel (NTK) of Matrix Product States (MPS) and the convergence of its NTK in the infinite bond dimensional limit. We prove that the NTK of MPS asymptotically converges to a constant matrix during the gradient descent (training) process (and also the initialization phase) as the bond dimensions of MPS go to infinity by the observation that the variation of the tensors in MPS asymptotically goes to zero during training in the infinite limit. By showing the positive-definiteness of the NTK of MPS, the convergence of MPS during the training in the function space (space of functions represented by MPS) is guaranteed without any extra assumptions of the data set. We then consider the settings of (supervised) Regression with Mean Square Error (RMSE) and (unsupervised) Born Machines (BM) and analyze their dynamics in the infinite bond dimensional limit. The ordinary differential equations (ODEs) which describe the dynamics of the responses of MPS in the RMSE and BM are derived and solved in the closed-form. For the Regression, we consider Mercer Kernels (Gaussian Kernels) and find that the evolution of the mean of the responses of MPS follows the largest eigenvalue of the NTK. Due to the orthogonality of the kernel functions in BM, the evolution of different modes (samples) decouples and the "characteristic time" of convergence in training is obtained.

stat.ML↗
arXiv2021

The Practical Scope of the Central Limit Theorem

The \textit{Central Limit Theorem (CLT)} is at the heart of a great deal of applied problem-solving in statistics and data science, but the theorem is silent on an important implementation issue: \textit{how much data do you need for the CLT to give accurate answers to practical questions?} Here we examine several approaches to addressing this issue -- along the way reviewing the history of this problem over the last 290 years -- and we illustrate the calculations with case-studies from finite-population sampling and gambling. A variety of surprises emerge.

stat.OT↗
arXiv2021

Representation Theorem for Matrix Product States

In this work, we investigate the universal representation capacity of the Matrix Product States (MPS) from the perspective of boolean functions and continuous functions. We show that MPS can accurately realize arbitrary boolean functions by providing a construction method of the corresponding MPS structure for an arbitrarily given boolean gate. Moreover, we prove that the function space of MPS with the scale-invariant sigmoidal activation is dense in the space of continuous functions defined on a compact subspace of the $n$-dimensional real coordinate space $\mathbb{R^{n}}$. We study the relation between MPS and neural networks and show that the MPS with a scale-invariant sigmoidal function is equivalent to a one-hidden-layer neural network equipped with a kernel function. We construct the equivalent neural networks for several specific MPS models and show that non-linear kernels such as the polynomial kernel which introduces the couplings between different components of the input into the model appear naturally in the equivalent neural networks. At last, we discuss the realization of the Gaussian Process (GP) with infinitely wide MPS by studying their equivalent neural networks.

stat.ML↗
arXiv2021

Infinitely Wide Tensor Networks as Gaussian Process

Gaussian Process is a non-parametric prior which can be understood as a distribution on the function space intuitively. It is known that by introducing appropriate prior to the weights of the neural networks, Gaussian Process can be obtained by taking the infinite-width limit of the Bayesian neural networks from a Bayesian perspective. In this paper, we explore the infinitely wide Tensor Networks and show the equivalence of the infinitely wide Tensor Networks and the Gaussian Process. We study the pure Tensor Network and another two extended Tensor Network structures: Neural Kernel Tensor Network and Tensor Network hidden layer Neural Network and prove that each one will converge to the Gaussian Process as the width of each model goes to infinity. (We note here that Gaussian Process can also be obtained by taking the infinite limit of at least one of the bond dimensions $\alpha_{i}$ in the product of tensor nodes, and the proofs can be done with the same ideas in the proofs of the infinite-width cases.) We calculate the mean function (mean vector) and the covariance function (covariance matrix) of the finite dimensional distribution of the induced Gaussian Process by the infinite-width tensor network with a general set-up. We study the properties of the covariance function and derive the approximation of the covariance function when the integral in the expectation operator is intractable. In the numerical experiments, we implement the Gaussian Process corresponding to the infinite limit tensor networks and plot the sample paths of these models. We study the hyperparameters and plot the sample path families in the induced Gaussian Process by varying the standard deviations of the prior distributions. As expected, the parameters in the prior distribution namely the hyper-parameters in the induced Gaussian Process controls the characteristic lengthscales of the Gaussian Process.

stat.ML↗
arXiv2021

The Bayesian Method of Tensor Networks

Bayesian learning is a powerful learning framework which combines the external information of the data (background information) with the internal information (training data) in a logically consistent way in inference and prediction. By Bayes rule, the external information (prior distribution) and the internal information (training data likelihood) are combined coherently, and the posterior distribution and the posterior predictive (marginal) distribution obtained by Bayes rule summarize the total information needed in the inference and prediction, respectively. In this paper, we study the Bayesian framework of the Tensor Network from two perspective. First, we introduce the prior distribution to the weights in the Tensor Network and predict the labels of the new observations by the posterior predictive (marginal) distribution. Since the intractability of the parameter integral in the normalization constant computation, we approximate the posterior predictive distribution by Laplace approximation and obtain the out-product approximation of the hessian matrix of the posterior distribution of the Tensor Network model. Second, to estimate the parameters of the stationary mode, we propose a stable initialization trick to accelerate the inference process by which the Tensor Network can converge to the stationary path more efficiently and stably with gradient descent method. We verify our work on the MNIST, Phishing Website and Breast Cancer data set. We study the Bayesian properties of the Bayesian Tensor Network by visualizing the parameters of the model and the decision boundaries in the two dimensional synthetic data set. For a application purpose, our work can reduce the overfitting and improve the performance of normal Tensor Network model.

stat.ML↗
arXiv2017

Comment: A brief survey of the current state of play for Bayesian computation in data science at Big-Data scale

We wish to contribute to the discussion of "Comparing Consensus Monte Carlo Strategies for Distributed Bayesian Computation" by offering our views on the current best methods for Bayesian computation, both at big-data scale and with smaller data sets, as summarized in Table 1. This table is certainly an over-simplification of a highly complicated area of research in constant (present and likely future) flux, but we believe that constructing summaries of this type is worthwhile despite their drawbacks, if only to facilitate further discussion.

stat.CO↗
arXiv2017

P\'olya Urn Latent Dirichlet Allocation: a doubly sparse massively parallel sampler

Latent Dirichlet Allocation (LDA) is a topic model widely used in natural language processing and machine learning. Most approaches to training the model rely on iterative algorithms, which makes it difficult to run LDA on big corpora that are best analyzed in parallel and distributed computational environments. Indeed, current approaches to parallel inference either don't converge to the correct posterior or require storage of large dense matrices in memory. We present a novel sampler that overcomes both problems, and we show that this sampler is faster, both empirically and theoretically, than previous Gibbs samplers for LDA. We do so by employing a novel P\'olya-urn-based approximation in the sparse partially collapsed sampler for LDA. We prove that the approximation error vanishes with data size, making our algorithm asymptotically exact, a property of importance for large-scale topic models. In addition, we show, via an explicit example, that - contrary to popular belief in the topic modeling literature - partially collapsed samplers can be more efficient than fully collapsed samplers. We conclude by comparing the performance of our algorithm with that of other approaches on well-known corpora.

stat.ML↗
arXiv2017

A Space-Based Observational Strategy for Characterizing the First Stars and Galaxies Using the Redshifted 21-cm Global Spectrum

The redshifted 21-cm monopole is expected to be a powerful probe of the epoch of the first stars and galaxies ($10<z<35$). The global 21-cm signal is sensitive to the thermal and ionization state of hydrogen gas and thus provides a tracer of sources of energetic photons -- primarily hot stars and accreting black holes -- which ionize and heat the high redshift intergalactic medium (IGM). This paper presents a strategy for observations of the global spectrum with a realizable instrument placed in a low altitude lunar orbit, performing night-time 40-120 MHz spectral observations, while on the farside to avoid terrestrial radio frequency interference, ionospheric corruption, and solar radio emissions. The frequency structure, uniformity over large scales, and unpolarized state of the redshifted 21-cm spectrum are distinct from the spectrally featureless, spatially-varying, and polarized emission from the bright foregrounds. This allows a clean separation between the primordial signal and foregrounds. For signal extraction, we model the foreground, instrument, and 21-cm spectrum with eigenmodes calculated via Singular Value Decomposition analyses. Using a Markov Chain Monte Carlo algorithm to explore the parameter space defined by the coefficients associated with these modes, we illustrate how the spectrum can be measured and how astrophysical parameters (e.g., IGM properties, first star characteristics) can be constrained in the presence of foregrounds using the Dark Ages Radio Explorer (DARE).

astro-ph.IM↗
arXiv2017

A Noninformative Prior on a Space of Distribution Functions

In a given problem, the Bayesian statistical paradigm requires the specification of a prior distribution that quantifies relevant information about the unknowns of main interest external to the data. In cases where little such information is available, the problem under study may possess an invariance under a transformation group that encodes a lack of information, leading to a unique prior---this idea was explored at length by E.T. Jaynes. Previous successful examples have included location-scale invariance under linear transformation, multiplicative invariance of the rate at which events in a counting process are observed, and the derivation of the Haldane prior for a Bernoulli success probability. In this paper we show that this method can be extended, by generalizing Jaynes, in two ways: (1) to yield families of approximately invariant priors, and (2) to the infinite-dimensional setting, yielding families of priors on spaces of distribution functions. Our results can be used to describe conditions under which a particular Dirichlet Process posterior arises from an optimal Bayesian analysis, in the sense that invariances in the prior and likelihood lead to one and only one posterior distribution.

math.ST↗
arXiv2016

GPU-accelerated Gibbs sampling: a case study of the Horseshoe Probit model

Gibbs sampling is a widely used Markov chain Monte Carlo (MCMC) method for numerically approximating integrals of interest in Bayesian statistics and other mathematical sciences. Many implementations of MCMC methods do not extend easily to parallel computing environments, as their inherently sequential nature incurs a large synchronization cost. In the case study illustrated by this paper, we show how to do Gibbs sampling in a fully data-parallel manner on a graphics processing unit, for a large class of exchangeable models that admit latent variable representations. Our approach takes a systems perspective, with emphasis placed on efficient use of compute hardware. We demonstrate our method on a Horseshoe Probit regression model and find that our implementation scales effectively to thousands of predictors and millions of data points simultaneously.

stat.CO↗
arXiv2015

Asynchronous Gibbs Sampling

Gibbs sampling is a Markov Chain Monte Carlo (MCMC) method often used in Bayesian learning. MCMC methods can be difficult to deploy on parallel and distributed systems due to their inherently sequential nature. We study asynchronous Gibbs sampling, which achieves parallelism by simply ignoring sequential requirements. This method has been shown to produce good empirical results for some hierarchical models, and is popular in the topic modeling community, but was also shown to diverge for other targets. We introduce a theoretical framework for analyzing asynchronous Gibbs sampling and other extensions of MCMC that do not possess the Markov property. We prove that asynchronous Gibbs can be modified so that it converges under appropriate regularity conditions -- we call this the exact asynchronous Gibbs algorithm. We study asynchronous Gibbs on a set of examples by comparing the exact and approximate algorithms, including two where it works well, and one where it fails dramatically. We conclude with a set of heuristics to describe settings where the algorithm can be effectively used.

stat.CO↗
arXiv2015

Causal Inference in Repeated Observational Studies: A Case Study of eBay Product Releases

Causal inference in observational studies is notoriously difficult, due to the fact that the experimenter is not in charge of the treatment assignment mechanism. Many potential con- founding factors (PCFs) exist in such a scenario, and if one seeks to estimate the causal effect of the treatment on a response, one needs to control for such factors. Identifying all relevant PCFs may be difficult (or impossible) given a single observational study. Instead, we argue that if one can observe a sequence of similar treatments over the course of a lengthy time period, one can identify patterns of behavior in the experimental subjects that are correlated with the response of interest and control for those patterns directly. Specifically, in our case-study we find and control for an early-adopter effect: the scenario in which the magnitude of the response is highly correlated with how quickly one adopts a treatment after its release. We provide a flexible hierarchical Bayesian framework that controls for such early-adopter effects in the analysis of the effects of multiple sequential treatments. The methods are presented and evaluated in the context of a detailed case-study involving product updates (newer versions of the same product) from eBay, Inc. The users in our study upgrade (or not) to a new version of the product at their own volition and timing. Our response variable is a measure of user actions, and we study the behavior of a large set of users (n = 10.5 million) in a targeted subset of eBay categories over a period of one year. We find that (a) naive causal estimates are hugely misleading and (b) our method, which is relatively insensitive to modeling assumptions and exhibits good out-of-sample predictive validation, yields sensible causal estimates that offer eBay a stable basis for decision-making.

stat.AP↗
arXiv2015

Cox's Theorem and the Jaynesian Interpretation of Probability

There are multiple proposed interpretations of probability theory: one such interpretation is true-false logic under uncertainty. Cox's Theorem is a representation theorem that states, under a certain set of axioms describing the meaning of uncertainty, that every true-false logic under uncertainty is isomorphic to conditional probability theory. This result was used by Jaynes to develop a philosophical framework in which statistical inference under uncertainty should be conducted through the use of probability, via Bayes' Rule. Unfortunately, most existing correct proofs of Cox's Theorem require restrictive assumptions: for instance, many do not apply even to the simple example of rolling a pair of fair dice. We offer a new axiomatization by replacing various technical conditions with an axiom stating that our theory must be consistent with respect to repeated events. We discuss the implications of our results, both for the philosophy of probability and for the philosophy of statistics.

math.ST↗
arXiv2014

A nonparametric Bayesian analysis of heterogeneous treatment effects in digital experimentation

Randomized controlled trials play an important role in how Internet companies predict the impact of policy decisions and product changes. In these `digital experiments', different units (people, devices, products) respond differently to the treatment. This article presents a fast and scalable Bayesian nonparametric analysis of such heterogeneous treatment effects and their measurement in relation to observable covariates. New results and algorithms are provided for quantifying the uncertainty associated with treatment effect measurement via both linear projections and nonlinear regression trees (CART and Random Forests). For linear projections, our inference strategy leads to results that are mostly in agreement with those from the frequentist literature. We find that linear regression adjustment of treatment effect averages (i.e., post-stratification) can provide some variance reduction, but that this reduction will be vanishingly small in the low-signal and large-sample setting of digital experiments. For regression trees, we provide uncertainty quantification for the machine learning algorithms that are commonly applied in tree-fitting. We argue that practitioners should look to ensembles of trees (forests) rather than individual trees in their analysis. The ideas are applied on and illustrated through an example experiment involving 21 million unique users of EBay.com.

stat.AP↗