arXiv subjects

Joong-Ho Won

Publications and source records attributed to Joong-Ho Won.

18 recordsLinked to original sources

arXiv2025

Learning Massive-scale Partial Correlation Networks in Clinical Multi-omics Studies with HP-ACCORD

Graphical model estimation from multi-omics data requires a balance between statistical estimation performance and computational scalability. We introduce a novel pseudolikelihood-based graphical model framework that reparameterizes the target precision matrix while preserving the sparsity pattern and estimates it by minimizing an $\ell_1$-penalized empirical risk based on a new loss function. The proposed estimator maintains estimation and selection consistency in various metrics under high-dimensional assumptions. The associated optimization problem allows for a provably fast computation algorithm using a novel operator-splitting approach and communication-avoiding distributed matrix multiplication. A high-performance computing implementation of our framework was tested using simulated data with up to one million variables, demonstrating complex dependency structures similar to those found in biological networks. Leveraging this scalability, we estimated a partial correlation network from a dual-omic liver cancer data set. The co-expression network estimated from the ultrahigh-dimensional data demonstrated superior specificity in prioritizing key transcription factors and co-activators by excluding the impact of epigenetic regulation, thereby highlighting the value of computational scalability in multi-omic data analysis.

stat.ML↗

arXiv2024

Discrete Consensus-Based Optimization

We propose Discrete Consensus-Based Optimization (DCBO), a fully discrete version of the Consensus-Based Optimization (CBO) framework. DCBO is a multi-agent method for the global optimization of possibly non-convex and non-differentiable functions. It aligns with the CBO paradigm, which promotes a consensus among agents towards a global optimum through simple stochastic dynamics amenable to rigorous mathematical analysis. Despite the promises, there has been a gap between the analysis of CBO and the actual behavior of the agents from its time-discrete implementation, as the former has focused on the system of continuous stochastic differential equations defining the model or its mean-field approximation. In particular, direct analysis of CBO-type algorithms with heterogeneous stochasticity is very challenging. DCBO distinguishes itself from these approaches in the sense that it has no continuous counterpart, thanks to the replacement of the "softmin" operator with the "hardmin" one, which is inherently discrete. Yet, it maintains the operational principles of CBO and allows for rich mathematical analysis. We present conditions, independent of the number of agents, for achieving a consensus or convergence and study the circumstances under which global optimization occurs. We test DCBO on a large number of benchmark functions to show its merits. We also demonstrate that DCBO is applicable to a diverse range of real-world problems, including neural network training, compressed sensing, and portfolio optimization, with competitive performance.

math.OC↗

arXiv2024

$t^3$-Variational Autoencoder: Learning Heavy-tailed Data with Student's t and Power Divergence

The variational autoencoder (VAE) typically employs a standard normal prior as a regularizer for the probabilistic latent encoder. However, the Gaussian tail often decays too quickly to effectively accommodate the encoded points, failing to preserve crucial structures hidden in the data. In this paper, we explore the use of heavy-tailed models to combat over-regularization. Drawing upon insights from information geometry, we propose $t^3$VAE, a modified VAE framework that incorporates Student's t-distributions for the prior, encoder, and decoder. This results in a joint model distribution of a power form which we argue can better fit real-world datasets. We derive a new objective by reformulating the evidence lower bound as joint optimization of KL divergence between two statistical manifolds and replacing with $γ$-power divergence, a natural alternative for power families. $t^3$VAE demonstrates superior generation of low-density regions when trained on heavy-tailed synthetic data. Furthermore, we show that $t^3$VAE significantly outperforms other models on CelebA and imbalanced CIFAR-100 datasets.

stat.ML↗

arXiv2024

On the Correctness of the Generalized Isotonic Recursive Partitioning Algorithm

This paper presents an in-depth analysis of the generalized isotonic recursive partitioning (GIRP) algorithm for fitting isotonic models under separable convex losses, proposed by Luss and Rosset [J. Comput. Graph. Statist., 23 (2014), pp. 192--201] for differentiable losses and extended by Painsky and Rosset [IEEE Trans. Pattern Anal. Mach. Intell., 38 (2016), pp. 308-321] for nondifferentiable losses. The GIRP algorithm poseses an attractive feature that in each step of the algorithm, the intermediate solution satisfies the isotonicity constraint. The paper begins with an example showing that the GIRP algorithm as described in the literature may fail to produce an isotonic model, suggesting that the existence and uniqueness of the solution to the isotonic regression problem must be carefully addressed. It proceeds with showing that, among possibly many solutions, there indeed exists a solution that can be found by recursive binary partitioning of the set of observed data. A small modification of the GIRP algorithm suffices to obtain a correct solution and preserve the desired property that all the intermediate solutions are isotonic. This proposed modification includes a proper choice of intermediate solutions and a simplification of the partitioning step from ternary to binary.

stat.ML↗

arXiv2023

Wasserstein Geodesic Generator for Conditional Distributions

Generating samples given a specific label requires estimating conditional distributions. We derive a tractable upper bound of the Wasserstein distance between conditional distributions to lay the theoretical groundwork to learn conditional distributions. Based on this result, we propose a novel conditional generation algorithm where conditional distributions are fully characterized by a metric space defined by a statistical distance. We employ optimal transport theory to propose the Wasserstein geodesic generator, a new conditional generator that learns the Wasserstein geodesic. The proposed method learns both conditional distributions for observed domains and optimal transport maps between them. The conditional distributions given unobserved intermediate domains are on the Wasserstein geodesic between conditional distributions given two observed domain labels. Experiments on face images with light conditions as domain labels demonstrate the efficacy of the proposed method.

stat.ML↗

arXiv2022

Statistical inference with implicit SGD: proximal Robbins-Monro vs. Polyak-Ruppert

The implicit stochastic gradient descent (ISGD), a proximal version of SGD, is gaining interest in the literature due to its stability over (explicit) SGD. In this paper, we conduct an in-depth analysis of the two modes of ISGD for smooth convex functions, namely proximal Robbins-Monro (proxRM) and proximal Poylak-Ruppert (proxPR) procedures, for their use in statistical inference on model parameters. Specifically, we derive non-asymptotic point estimation error bounds of both proxRM and proxPR iterates and their limiting distributions, and propose on-line estimators of their asymptotic covariance matrices that require only a single run of ISGD. The latter estimators are used to construct valid confidence intervals for the model parameters. Our analysis is free of the generalized linear model assumption that has limited the preceding analyses, and employs feasible procedures. Our on-line covariance matrix estimators appear to be the first of this kind in the ISGD literature.

stat.ML↗

arXiv2022

A unified analysis of convex and non-convex lp-ball projection problems

The task of projecting onto $\ell_p$ norm balls is ubiquitous in statistics and machine learning, yet the availability of actionable algorithms for doing so is largely limited to the special cases of $p = \left\{ 0, 1,2, \infty \right\}$. In this paper, we introduce novel, scalable methods for projecting onto the $\ell_p$ ball for general $p>0$. For $p \geq1 $, we solve the univariate Lagrangian dual via a dual Newton method. We then carefully design a bisection approach for $p<1$, presenting theoretical and empirical evidence of zero or a small duality gap in the non-convex case. The success of our contributions is thoroughly assessed empirically, and applied to large-scale regularized multi-task learning and compressed sensing.

math.OC↗

arXiv2021

Orthogonal Trace-Sum Maximization: Tightness of the Semidefinite Relaxation and Guarantee of Locally Optimal Solutions

This paper studies an optimization problem on the sum of traces of matrix quadratic forms in $m$ semi-orthogonal matrices, which can be considered as a generalization of the synchronization of rotations. While the problem is nonconvex, the paper shows that its semidefinite programming relaxation solves the original nonconvex problems exactly with high probability, under an additive noise model with small noise in the order of $O(m^{1/4})$. In addition, it shows that the solution of a nonconvex algorithm considered in Won, Zhou, and Lange [SIAM J. Matrix Anal. Appl., 2 (2021), pp. 859-882] is also its global solution with high probability under similar conditions. These results can be considered as a generalization of existing results on phase synchronization.

math.OC↗

arXiv2021

High-Performance Statistical Computing in the Computing Environments of the 2020s

Technological advances in the past decade, hardware and software alike, have made access to high-performance computing (HPC) easier than ever. We review these advances from a statistical computing perspective. Cloud computing makes access to supercomputers affordable. Deep learning software libraries make programming statistical algorithms easy and enable users to write code once and run it anywhere -- from a laptop to a workstation with multiple graphics processing units (GPUs) or a supercomputer in a cloud. Highlighting how these developments benefit statisticians, we review recent optimization algorithms that are useful for high-dimensional models and can harness the power of HPC. Code snippets are provided to demonstrate the ease of programming. We also provide an easy-to-use distributed matrix data structure suitable for HPC. Employing this data structure, we illustrate various statistical applications including large-scale positron emission tomography and $\ell_1$-regularized Cox regression. Our examples easily scale up to an 8-GPU workstation and a 720-CPU-core cluster in a cloud. As a case in point, we analyze the onset of type-2 diabetes from the UK Biobank with 200,000 subjects and about 500,000 single nucleotide polymorphisms using the HPC $\ell_1$-regularized Cox regression. Fitting this half-million-variate model takes less than 45 minutes and reconfirms known associations. To our knowledge, this is the first demonstration of the feasibility of penalized regression of survival outcomes at this scale.

stat.CO↗

arXiv2021

Nonconvex Optimization via MM Algorithms: Convergence Theory

The majorization-minimization (MM) principle is an extremely general framework for deriving optimization algorithms. It includes the expectation-maximization (EM) algorithm, proximal gradient algorithm, concave-convex procedure, quadratic lower bound algorithm, and proximal distance algorithm as special cases. Besides numerous applications in statistics, optimization, and imaging, the MM principle finds wide applications in large scale machine learning problems such as matrix completion, discriminant analysis, and nonnegative matrix factorizations. When applied to nonconvex optimization problems, MM algorithms enjoy the advantages of convexifying the objective function, separating variables, numerical stability, and ease of implementation. However, compared to the large body of literature on other optimization algorithms, the convergence analysis of MM algorithms is scattered and problem specific. This survey presents a unified treatment of the convergence of MM algorithms. With modern applications in mind, the results encompass non-smooth objective functions and non-asymptotic analysis.

math.OC↗

arXiv2021

Orthogonal Trace-Sum Maximization: Applications, Local Algorithms, and Global Optimality

This paper studies the problem of maximizing the sum of traces of matrix quadratic forms on a product of Stiefel manifolds. This orthogonal trace-sum maximization (OTSM) problem generalizes many interesting problems such as generalized canonical correlation analysis (CCA), Procrustes analysis, and cryo-electron microscopy of the Nobel prize fame. For these applications finding global solutions is highly desirable but it has been unclear how to find even a stationary point, let alone testing its global optimality. Through a close inspection of Ky Fan's classical result (1949) on the variational formulation of the sum of largest eigenvalues of a symmetric matrix, and a semidefinite programming (SDP) relaxation of the latter, we first provide a simple method to certify global optimality of a given stationary point of OTSM. This method only requires testing whether a symmetric matrix is positive semidefinite. A by-product of this analysis is an unexpected strong duality between Shapiro-Botha (1988) and Zhang-Singer (2017). After showing that a popular algorithm for generalized CCA and Procrustes analysis may generate oscillating iterates, we propose a simple fix that provably guarantees convergence to a stationary point. The combination of our algorithm and certificate reveals novel global optima of various instances of OTSM.

math.OC↗

arXiv2020

DistStat.jl: Towards Unified Programming for High-Performance Statistical Computing Environments in Julia

The demand for high-performance computing (HPC) is ever-increasing for everyday statistical computing purposes. The downside is that we need to write specialized code for each HPC environment. CPU-level parallelization needs to be explicitly coded for effective use of multiple nodes in cluster supercomputing environments. Acceleration via graphics processing units (GPUs) requires to write kernel code. The Julia software package DistStat.jl implements a data structure for distributed arrays that work on both multi-node CPU clusters and multi-GPU environments transparently. This package paves a way to developing high-performance statistical software in various HPC environments simultaneously. As a demonstration of the transparency and scalability of the package, we provide applications to large-scale nonnegative matrix factorization, multidimensional scaling, and $\ell_1$-regularized Cox proportional hazards model on an 8-GPU workstation and a 720-CPU-core virtual cluster in Amazon Web Services (AWS) cloud. As a case in point, we analyze the on-set of type-2 diabetes from the UK Biobank with 400,000 subjects and 500,000 single nucleotide polymorphisms using the $\ell_1$-regularized Cox proportional hazards model. Fitting a half-million-variate regression model took less than 50 minutes on AWS.

stat.CO↗

arXiv2020

Principled learning method for Wasserstein distributionally robust optimization with local perturbations

Wasserstein distributionally robust optimization (WDRO) attempts to learn a model that minimizes the local worst-case risk in the vicinity of the empirical data distribution defined by Wasserstein ball. While WDRO has received attention as a promising tool for inference since its introduction, its theoretical understanding has not been fully matured. Gao et al. (2017) proposed a minimizer based on a tractable approximation of the local worst-case risk, but without showing risk consistency. In this paper, we propose a minimizer based on a novel approximation theorem and provide the corresponding risk consistency results. Furthermore, we develop WDRO inference for locally perturbed data that include the Mixup (Zhang et al., 2017) as a special case. We show that our approximation and risk consistency results naturally extend to the cases when data are locally perturbed. Numerical experiments demonstrate robustness of the proposed method using image classification datasets. Our results show that the proposed method achieves significantly higher accuracy than baseline models on noisy datasets.

stat.ML↗

arXiv2019

Splitting with Near-Circulant Linear Systems: Applications to Total Variation CT and PET

Many imaging problems, such as total variation reconstruction of X-ray computed tomography (CT) and positron-emission tomography (PET), are solved via a convex optimization problem with near-circulant, but not actually circulant, linear systems. The popular methods to solve these problems, alternating direction method of multipliers (ADMM) and primal-dual hybrid gradient (PDHG), do not directly utilize this structure. Consequently, ADMM requires a costly matrix inversion as a subroutine, and PDHG takes too many iterations to converge. In this paper, we present near-circulant splitting (NCS), a novel splitting method that leverages the near-circulant structure. We show that NCS can converge with an iteration count close to that of ADMM, while paying a computational cost per iteration close to that of PDHG. Through experiments on a CUDA GPU, we empirically validate the theory and demonstrate that NCS can effectively utilize the parallel computing capabilities of CUDA.

math.OC↗

arXiv2018

Easily parallelizable and distributable class of algorithms for structured sparsity, with optimal acceleration

Many statistical learning problems can be posed as minimization of a sum of two convex functions, one typically a composition of non-smooth and linear functions. Examples include regression under structured sparsity assumptions. Popular algorithms for solving such problems, e.g., ADMM, often involve non-trivial optimization subproblems or smoothing approximation. We consider two classes of primal-dual algorithms that do not incur these difficulties, and unify them from a perspective of monotone operator theory. From this unification we propose a continuum of preconditioned forward-backward operator splitting algorithms amenable to parallel and distributed computing. For the entire region of convergence of the whole continuum of algorithms, we establish its rates of convergence. For some known instances of this continuum, our analysis closes the gap in theory. We further exploit the unification to propose a continuum of accelerated algorithms. We show that the whole continuum attains the theoretically optimal rate of convergence. The scalability of the proposed algorithms, as well as their convergence behavior, is demonstrated up to 1.2 million variables with a distributed implementation.

stat.ML↗

arXiv2016

Towards a sparse, scalable, and stably positive definite (inverse) covariance estimator

High dimensional covariance estimation and graphical models is a contemporary topic in statistics and machine learning having widespread applications. An important line of research in this regard is to shrink the extreme spectrum of the covariance matrix estimators. A separate line of research in the literature has considered sparse inverse covariance estimation which in turn gives rise to graphical models. In practice, however, a sparse covariance or inverse covariance matrix which is simultaneously well-conditioned and at the same time computationally tractable is desired. There has been little research at the confluence of these three topics. In this paper we consider imposing a condition number constraint to various types of losses used in covariance and inverse covariance matrix estimation. When the loss function can be decomposed as a sum of an orthogonally invariant function of the estimate and its inner product with a function of the sample covariance matrix, we show that a solution path algorithm can be derived, involving a series of ordinary differential equations. The path algorithm is attractive because it provides the entire family of estimates for all possible values of the condition number bound, at the same computational cost of a single estimate with a fixed upper bound. An important finding is that the proximal operator for the condition number constraint, which turns out to be very useful in regularizing loss functions that are not orthogonally invariant and may yield non-positive-definite estimates, can be efficiently computed by this path algorithm. As a concrete illustration of its practical importance, we develop an operator-splitting algorithm that imposes a guarantee of well-conditioning as well as positive definiteness to recently proposed convex pseudo-likelihood based graphical model selection methods.

stat.ME↗

arXiv2013

High-dimensional Fused Lasso Regression using Majorization-Minimization and Parallel Processing

In this paper, we propose a majorization-minimization (MM) algorithm for high-dimensional fused lasso regression (FLR) suitable for parallelization using graphics processing units (GPUs). The MM algorithm is stable and flexible as it can solve the FLR problems with various types of design matrices and penalty structures within a few tens of iterations. We also show that the convergence of the proposed algorithm is guaranteed. We conduct numerical studies to compare our algorithm with other existing algorithms, demonstrating that the proposed MM algorithm is competitive in many settings including the two-dimensional FLR with arbitrary design matrices. The merit of GPU parallelization is also exhibited.

stat.ME↗

arXiv2013

Monotone false discovery rate

This paper proposes a procedure to obtain monotone estimates of both the local and the tail false discovery rates that arise in large-scale multiple testing. The proposed monotonization is asymptotically optimal for controlling the false discovery rate and also has many attractive finite-sample properties.

stat.ME↗