arXiv subjects

Arian Maleki

Publications and source records attributed to Arian Maleki.

At least 37 records · Page 2Linked to original sources

arXiv2024

Theoretical Analysis of Leave-one-out Cross Validation for Non-differentiable Penalties under High-dimensional Settings

Despite a large and significant body of recent work focused on estimating the out-of-sample risk of regularized models in the high dimensional regime, a theoretical understanding of this problem for non-differentiable penalties such as generalized LASSO and nuclear norm is missing. In this paper we resolve this challenge. We study this problem in the proportional high dimensional regime where both the sample size n and number of features p are large, and n/p and the signal-to-noise ratio (per observation) remain finite. We provide finite sample upper bounds on the expected squared error of leave-one-out cross-validation (LO) in estimating the out-of-sample risk. The theoretical framework presented here provides a solid foundation for elucidating empirical findings that show the accuracy of LO.

math.ST↗

arXiv2023

Signal-to-noise ratio aware minimaxity and higher-order asymptotics

Since its development, the minimax framework has been one of the corner stones of theoretical statistics, and has contributed to the popularity of many well-known estimators, such as the regularized M-estimators for high-dimensional problems. In this paper, we will first show through the example of sparse Gaussian sequence model, that the theoretical results under the classical minimax framework are insufficient for explaining empirical observations. In particular, both hard and soft thresholding estimators are (asymptotically) minimax, however, in practice they often exhibit sub-optimal performances at various signal-to-noise ratio (SNR) levels. The first contribution of this paper is to demonstrate that this issue can be resolved if the signal-to-noise ratio is taken into account in the construction of the parameter space. We call the resulting minimax framework the signal-to-noise ratio aware minimaxity. The second contribution of this paper is to showcase how one can use higher-order asymptotics to obtain accurate approximations of the SNR-aware minimax risk and discover minimax estimators. The theoretical findings obtained from this refined minimax framework provide new insights and practical guidance for the estimation of sparse signals.

math.ST↗

arXiv2023

Approximate Leave-one-out Cross Validation for Regression with $\ell_1$ Regularizers (extended version)

The out-of-sample error (OO) is the main quantity of interest in risk estimation and model selection. Leave-one-out cross validation (LO) offers a (nearly) distribution-free yet computationally demanding approach to estimate OO. Recent theoretical work showed that approximate leave-one-out cross validation (ALO) is a computationally efficient and statistically reliable estimate of LO (and OO) for generalized linear models with differentiable regularizers. For problems involving non-differentiable regularizers, despite significant empirical evidence, the theoretical understanding of ALO's error remains unknown. In this paper, we present a novel theory for a wide class of problems in the generalized linear model family with non-differentiable regularizers. We bound the error |ALO - LO| in terms of intuitive metrics such as the size of leave-i-out perturbations in active sets, sample size n, number of features p and regularization parameters. As a consequence, for the $\ell_1$-regularized problems, we show that |ALO - LO| goes to zero as p goes to infinity while n/p and SNR are fixed and bounded.

math.ST↗

arXiv2023

Towards Designing Optimal Sensing Matrices for Generalized Linear Inverse Problems

We consider an inverse problem $\mathbf{y}= f(\mathbf{Ax})$, where $\mathbf{x}\in\mathbb{R}^n$ is the signal of interest, $\mathbf{A}$ is the sensing matrix, $f$ is a nonlinear function and $\mathbf{y} \in \mathbb{R}^m$ is the measurement vector. In many applications, we have some level of freedom to design the sensing matrix $\mathbf{A}$, and in such circumstances we could optimize $\mathbf{A}$ to achieve better reconstruction performance. As a first step towards optimal design, it is important to understand the impact of the sensing matrix on the difficulty of recovering $\mathbf{x}$ from $\mathbf{y}$. In this paper, we study the performance of one of the most successful recovery methods, i.e., the expectation propagation (EP) algorithm. We define a notion of spikiness for the spectrum of $\bmmathbfA}$ and show the importance of this measure for the performance of EP. We show that whether a spikier spectrum can hurt or help the recovery performance depends on $f$. Based on our framework, we are able to show that, in phase-retrieval problems, matrices with spikier spectrums are better for EP, while in 1-bit compressed sensing problems, less spiky spectrums lead to better performance. Our results unify and substantially generalize existing results that compare Gaussian and orthogonal matrices, and provide a platform towards designing optimal sensing systems.

cs.IT↗

arXiv2022

Sharp Concentration Results for Heavy-Tailed Distributions

We obtain concentration and large deviation for the sums of independent and identically distributed random variables with heavy-tailed distributions. Our concentration results are concerned with random variables whose distributions satisfy $\mathbb{P}(X>t) \leq {\rm e}^{- I(t)}$, where $I: \mathbb{R} \rightarrow \mathbb{R}$ is an increasing function and $I(t)/t \rightarrow α\in [0, \infty)$ as $t \rightarrow \infty$. Our main theorem can not only recover some of the existing results, such as the concentration of the sum of subWeibull random variables, but it can also produce new results for the sum of random variables with heavier tails. We show that the concentration inequalities we obtain are sharp enough to offer large deviation results for the sums of independent random variables as well. Our analyses which are based on standard truncation arguments simplify, unify and generalize the existing results on the concentration and large deviation of heavy-tailed random variables.

math.PR↗

arXiv2021

Mismatched Data Detection in Massive MU-MIMO

We investigate mismatched data detection for massive multi-user (MU) multiple-input multiple-output (MIMO) wireless systems in which the prior distribution of the transmit signal used in the data detector differs from the true prior. In order to minimize the performance loss caused by the prior mismatch, we include a tuning stage into the recently proposed large-MIMO approximate message passing (LAMA) algorithm, which enables the development of data detectors with optimal as well as sub-optimal parameter tuning. We show that carefully-selected priors enable the design of simpler and computationally more efficient data detection algorithms compared to LAMA that uses the optimal prior, while achieving near-optimal error-rate performance. In particular, we demonstrate that a hardware-friendly approximation of the exact prior enables the design of low-complexity data detectors that achieve near individually-optimal performance. Furthermore, for Gaussian priors and uniform priors within a hypercube covering the quadrature amplitude modulation (QAM) constellation, our performance analysis recovers classical and recent results on linear and non-linear massive MU-MIMO data detection, respectively.

cs.IT↗

arXiv2021

Does SLOPE outperform bridge regression?

A recently proposed SLOPE estimator (arXiv:1407.3824) has been shown to adaptively achieve the minimax $\ell_2$ estimation rate under high-dimensional sparse linear regression models (arXiv:1503.08393). Such minimax optimality holds in the regime where the sparsity level $k$, sample size $n$, and dimension $p$ satisfy $k/p \rightarrow 0$, $k\log p/n \rightarrow 0$. In this paper, we characterize the estimation error of SLOPE under the complementary regime where both $k$ and $n$ scale linearly with $p$, and provide new insights into the performance of SLOPE estimators. We first derive a concentration inequality for the finite sample mean square error (MSE) of SLOPE. The quantity that MSE concentrates around takes a complicated and implicit form. With delicate analysis of the quantity, we prove that among all SLOPE estimators, LASSO is optimal for estimating $k$-sparse parameter vectors that do not have tied non-zero components in the low noise scenario. On the other hand, in the large noise scenario, the family of SLOPE estimators are sub-optimal compared with bridge regression such as the Ridge estimator.

stat.ML↗

arXiv2021

Compressed sensing in the presence of speckle noise

The problem of recovering a structured signal from its linear measurements in the presence of speckle noise is studied. This problem appears in many imaging systems such as synthetic aperture radar and optical coherence tomography. The current acquisition technology oversamples signals and converts the problem into a denoising problem with multiplicative noise. However, this paper explores the possibility of reducing the number of measurements below the ambient dimension of the signal. The sophistications that appear in the study of multiplicative noises have so far impeded theoretical analysis of such problems. This paper aims to present the first theoretical result regarding the recovery of signals from their undersampled measurements under the speckle noise. It is shown that if the signal class is structured, in the sense that the signals can be compressed efficiently, then one can obtain accurate estimates of the signal from fewer measurements than the ambient dimension. We demonstrate the effectiveness of the methods we propose through simulation results.

cs.IT↗

arXiv2021

Minimax Linear Estimation of the Retargeted Mean

Evaluating treatments received by one population for application to a different target population of scientific interest is a central problem in causal inference from observational studies. We study the minimax linear estimator of the treatment-specific mean outcome on a target population and provide a theoretical basis for inference based on it. In particular, we provide a justification for the common practice of ignoring bias when building confidence intervals with these linear estimators. Focusing on the case that the class of the unknown outcome function is the unit ball of a reproducing kernel Hilbert space, we show that the resulting linear estimator is asymptotically optimal under conditions only marginally stronger than those used with augmented estimators. We establish bounds attesting to the estimator's good finite sample properties. In an extensive simulation study, we observe promising performance of the estimator throughout a wide range of sample sizes, noise levels, and levels of overlap between the covariate distributions of the treated and target populations.

math.ST↗

arXiv2021

Consistent Risk Estimation in Moderately High-Dimensional Linear Regression

Risk estimation is at the core of many learning systems. The importance of this problem has motivated researchers to propose different schemes, such as cross validation, generalized cross validation, and Bootstrap. The theoretical properties of such estimates have been extensively studied in the low-dimensional settings, where the number of predictors $p$ is much smaller than the number of observations $n$. However, a unifying methodology accompanied with a rigorous theory is lacking in high-dimensional settings. This paper studies the problem of risk estimation under the moderately high-dimensional asymptotic setting $n,p \rightarrow \infty$ and $n/p \rightarrow δ>1$ ($δ$ is a fixed number), and proves the consistency of three risk estimates that have been successful in numerical studies, i.e., leave-one-out cross validation (LOOCV), approximate leave-one-out (ALO), and approximate message passing (AMP)-based techniques. A corner stone of our analysis is a bound that we obtain on the discrepancy of the `residuals' obtained from AMP and LOOCV. This connection not only enables us to obtain a more refined information on the estimates of AMP, ALO, and LOOCV, but also offers an upper bound on the convergence rate of each estimate.

math.ST↗

arXiv2020

Spectral Method for Phase Retrieval: an Expectation Propagation Perspective

Phase retrieval refers to the problem of recovering a signal $\mathbf{x}_{\star}\in\mathbb{C}^n$ from its phaseless measurements $y_i=|\mathbf{a}_i^{\mathrm{H}}\mathbf{x}_{\star}|$, where $\{\mathbf{a}_i\}_{i=1}^m$ are the measurement vectors. Many popular phase retrieval algorithms are based on the following two-step procedure: (i) initialize the algorithm based on a spectral method, (ii) refine the initial estimate by a local search algorithm (e.g., gradient descent). The quality of the spectral initialization step can have a major impact on the performance of the overall algorithm. In this paper, we focus on the model where the measurement matrix $\mathbf{A}=[\mathbf{a}_1,\ldots,\mathbf{a}_m]^{\mathrm{H}}$ has orthonormal columns, and study the spectral initialization under the asymptotic setting $m,n\to\infty$ with $m/n\toδ\in(1,\infty)$. We use the expectation propagation framework to characterize the performance of spectral initialization for Haar distributed matrices. Our numerical results confirm that the predictions of the EP method are accurate for not-only Haar distributed matrices, but also for realistic Fourier based models (e.g. the coded diffraction model). The main findings of this paper are the following: (1) There exists a threshold on $δ$ (denoted as $δ_{\mathrm{weak}}$) below which the spectral method cannot produce a meaningful estimate. We show that $δ_{\mathrm{weak}}=2$ for the column-orthonormal model. In contrast, previous results by Mondelli and Montanari show that $δ_{\mathrm{weak}}=1$ for the i.i.d. Gaussian model. (2) The optimal design for the spectral method coincides with that for the i.i.d. Gaussian model, where the latter was recently introduced by Luo, Alghamdi and Lu.

cs.IT↗

arXiv2020

Optimal Data Detection and Signal Estimation in Systems with Input Noise

Practical systems often suffer from hardware impairments that already appear during signal generation. Despite the limiting effect of such input-noise impairments on signal processing systems, they are routinely ignored in the literature. In this paper, we propose an algorithm for data detection and signal estimation, referred to as Approximate Message Passing with Input noise (AMPI), which takes into account input-noise impairments. To demonstrate the efficacy of AMPI, we investigate two applications: Data detection in large multiple-input multiple output (MIMO) wireless systems and sparse signal recovery in compressive sensing. For both applications, we provide precise conditions in the large-system limit for which AMPI achieves optimal performance. We furthermore use simulations to demonstrate that AMPI achieves near-optimal performance at low complexity in realistic, finite-dimensional systems.

eess.SP↗

arXiv2020

Information Theoretic Limits for Phase Retrieval with Subsampled Haar Sensing Matrices

We study information theoretic limits of recovering an unknown $n$ dimensional, complex signal vector $\mathbf{x}_\star$ with unit norm from $m$ magnitude-only measurements of the form $y_i = |(\mathbf{A} \mathbf{x}_\star)_i|^2, \; i = 1,2 \dots , m$, where $\mathbf{A}$ is the sensing matrix. This is known as the Phase Retrieval problem and models practical imaging systems where measuring the phase of the observations is difficult. Since in a number of applications, the sensing matrix has orthogonal columns, we model the sensing matrix as a subsampled Haar matrix formed by picking $n$ columns of a uniformly random $m \times m$ unitary matrix. We study this problem in the high dimensional asymptotic regime, where $m,n \rightarrow \infty$, while $m/n \rightarrow δ$ with $δ$ being a fixed number, and show that if $m < (2-o_n(1))\cdot n$, then any estimator is asymptotically orthogonal to the true signal vector $\mathbf{x}_\star$. This lower bound is sharp since when $m > (2+o_n(1)) \cdot n $, estimators that achieve a non trivial asymptotic correlation with the signal vector are known from previous works.

math.ST↗

arXiv2020

Using Black-box Compression Algorithms for Phase Retrieval

Compressive phase retrieval refers to the problem of recovering a structured $n$-dimensional complex-valued vector from its phase-less under-determined linear measurements. The non-linearity of measurements makes designing theoretically-analyzable efficient phase retrieval algorithms challenging. As a result, to a great extent, algorithms designed in this area are developed to take advantage of simple structures such as sparsity and its convex generalizations. The goal of this paper is to move beyond simple models through employing compression codes. Such codes are typically developed to take advantage of complex signal models to represent the signals as efficiently as possible. In this work, it is shown how an existing compression code can be treated as a black box and integrated into an efficient solution for phase retrieval. First, COmpressive PhasE Retrieval (COPER) optimization, a computationally-intensive compression-based phase retrieval method, is proposed. COPER provides a theoretical framework for studying compression-based phase retrieval. The number of measurements required by COPER is connected to $κ$, the $α$-dimension (closely related to the rate-distortion dimension) of the given family of compression codes. To finds the solution of COPER, an efficient iterative algorithm called gradient descent for COPER (GD-COPER) is proposed. It is proven that under some mild conditions on the initialization, if the number of measurements is larger than $ C κ^2 \log^2 n$, where $C$ is a constant, GD-COPER obtains an accurate estimate of the input vector in polynomial time. In the simulation results, JPEG2000 is integrated in GD-COPER to confirm the superb performance of the resulting algorithm on real-world images.

cs.IT↗

arXiv2020

Which bridge estimator is optimal for variable selection?

We study the problem of variable selection for linear models under the high-dimensional asymptotic setting, where the number of observations $n$ grows at the same rate as the number of predictors $p$. We consider two-stage variable selection techniques (TVS) in which the first stage uses bridge estimators to obtain an estimate of the regression coefficients, and the second stage simply thresholds this estimate to select the "important" predictors. The asymptotic false discovery proportion (AFDP) and true positive proportion (ATPP) of these TVS are evaluated. We prove that for a fixed ATPP, in order to obtain a smaller AFDP, one should pick a bridge estimator with smaller asymptotic mean square error in the first stage of TVS. Based on such principled discovery, we present a sharp comparison of different TVS, via an in-depth investigation of the estimation properties of bridge estimators. Rather than "order-wise" error bounds with loose constants, our analysis focuses on precise error characterization. Various interesting signal-to-noise ratio and sparsity settings are studied. Our results offer new and thorough insights into high-dimensional variable selection. For instance, we prove that a TVS with Ridge in its first stage outperforms TVS with other bridge estimators in large noise settings; two-stage LASSO becomes inferior when the signal is rare and weak. As a by-product, we show that two-stage methods outperform some standard variable selection techniques, such as LASSO and Sure Independence Screening, under certain conditions.

math.ST↗

arXiv2020

Analysis of Spectral Methods for Phase Retrieval with Random Orthogonal Matrices

Phase retrieval refers to algorithmic methods for recovering a signal from its phaseless measurements. Local search algorithms that work directly on the non-convex formulation of the problem have been very popular recently. Due to the nonconvexity of the problem, the success of these local search algorithms depends heavily on their starting points. The most widely used initialization scheme is the spectral method, in which the leading eigenvector of a data-dependent matrix is used as a starting point. Recently, the performance of the spectral initialization was characterized accurately for measurement matrices with independent and identically distributed entries. This paper aims to obtain the same level of knowledge for isotropically random column-orthogonal matrices, which are substantially better models for practical phase retrieval systems. Towards this goal, we consider the asymptotic setting in which the number of measurements $m$, and the dimension of the signal, $n$, diverge to infinity with $m/n = δ\in(1,\infty)$, and obtain a simple expression for the overlap between the spectral estimator and the true signal vector.

cs.IT↗

arXiv2020

Error bounds in estimating the out-of-sample prediction error using leave-one-out cross validation in high-dimensions

We study the problem of out-of-sample risk estimation in the high dimensional regime where both the sample size $n$ and number of features $p$ are large, and $n/p$ can be less than one. Extensive empirical evidence confirms the accuracy of leave-one-out cross validation (LO) for out-of-sample risk estimation. Yet, a unifying theoretical evaluation of the accuracy of LO in high-dimensional problems has remained an open problem. This paper aims to fill this gap for penalized regression in the generalized linear family. With minor assumptions about the data generating process, and without any sparsity assumptions on the regression coefficients, our theoretical analysis obtains finite sample upper bounds on the expected squared error of LO in estimating the out-of-sample error. Our bounds show that the error goes to zero as $n,p \rightarrow \infty$, even when the dimension $p$ of the feature vectors is comparable with or greater than the sample size $n$. One technical advantage of the theory is that it can be used to clarify and connect some results from the recent literature on scalable approximate LO.

stat.ML↗

arXiv2020

A scalable estimate of the extra-sample prediction error via approximate leave-one-out

The paper considers the problem of out-of-sample risk estimation under the high dimensional settings where standard techniques such as $K$-fold cross validation suffer from large biases. Motivated by the low bias of the leave-one-out cross validation (LO) method, we propose a computationally efficient closed-form approximate leave-one-out formula (ALO) for a large class of regularized estimators. Given the regularized estimate, calculating ALO requires minor computational overhead. With minor assumptions about the data generating process, we obtain a finite-sample upper bound for $|\text{LO} - \text{ALO}|$. Our theoretical analysis illustrates that $|\text{LO} - \text{ALO}| \rightarrow 0$ with overwhelming probability, when $n,p \rightarrow \infty$, where the dimension $p$ of the feature vectors may be comparable with or even greater than the number of observations, $n$. Despite the high-dimensionality of the problem, our theoretical results do not require any sparsity assumption on the vector of regression coefficients. Our extensive numerical experiments show that $|\text{LO} - \text{ALO}|$ decreases as $n,p$ increase, revealing the excellent finite sample performance of ALO. We further illustrate the usefulness of our proposed out-of-sample risk estimation method by an example of real recordings from spatially sensitive neurons (grid cells) in the medial entorhinal cortex of a rat.

stat.ME↗