Sharp global convergence guarantees for iterative nonconvex optimization with random data
From MaRDI portal
Abstract: We consider a general class of regression models with normally distributed covariates, and the associated nonconvex problem of fitting these models from data. We develop a general recipe for analyzing the convergence of iterative algorithms for this task from a random initialization. In particular, provided each iteration can be written as the solution to a convex optimization problem satisfying some natural conditions, we leverage Gaussian comparison theorems to derive a deterministic sequence that provides sharp upper and lower bounds on the error of the algorithm with sample-splitting. Crucially, this deterministic sequence accurately captures both the convergence rate of the algorithm and the eventual error floor in the finite-sample regime, and is distinct from the commonly used "population" sequence that results from taking the infinite-sample limit. We apply our general framework to derive several concrete consequences for parameter estimation in popular statistical models including phase retrieval and mixtures of regressions. Provided the sample size scales near-linearly in the dimension, we show sharp global convergence rates for both higher-order algorithms based on alternating updates and first-order algorithms based on subgradient descent. These corollaries, in turn, yield multiple consequences, including: (a) Proof that higher-order algorithms can converge significantly faster than their first-order counterparts (and sometimes super-linearly), even if the two share the same population update and (b) Intricacies in super-linear convergence behavior for higher-order algorithms, which can be nonstandard (e.g., with exponent 3/2) and sensitive to the noise level in the problem. We complement these results with extensive numerical experiments, which show excellent agreement with our theoretical predictions.
Recommendations
- On the Global Convergence of Randomized Coordinate Gradient Descent for Nonconvex Optimization
- On Global Linear Convergence in Stochastic Nonconvex Optimization for Semidefinite Programming
- Convergence guarantees for a class of non-convex and non-smooth optimization problems
- Nonasymptotic convergence of stochastic proximal point methods for constrained convex optimization
- Global convergence of nonmonotone descent methods for unconstrained optimization problems
- Linear Convergence of Random Dual Coordinate Descent on Nonpolyhedral Convex Problems
- Convergence of the Gradient Sampling Algorithm for Nonsmooth Nonconvex Optimization
- Global convergence of a non-convex Douglas-Rachford iteration
- Convergence of simultaneous perturbation stochastic approximation for nondifferentiable optimization
- Large-Scale Nonconvex Optimization: Randomization, Gap Estimation, and Numerical Resolution
Cites work
- A model of double descent for high-dimensional binary linear classification
- A modern maximum-likelihood theory for high-dimensional logistic regression
- A nonconvex approach for phase retrieval: reshaped Wirtinger flow and incremental algorithms
- A precise high-dimensional asymptotic theory for boosting and minimum-\(\ell_1\)-norm interpolated classifiers
- Does SLOPE outperform bridge regression?
- Estimating the Coefficients of a Mixture of Two Linear Regressions by Expectation Maximization
- Fast and reliable parameter estimation from nonlinear observations
- Fast global convergence of gradient methods for high-dimensional statistical recovery
- Global Guarantees for Enforcing Deep Generative Priors by Empirical Risk
- Gradient descent with random initialization: fast global convergence for nonconvex phase retrieval
- Guaranteed Matrix Completion via Non-Convex Factorization
- High-dimensional probability. An introduction with applications in data science
- High-dimensional regression with noisy and missing data: provable guarantees with nonconvexity
- scientific article; zbMATH DE number 4061904 (Why is no real title available?)
- scientific article; zbMATH DE number 3567782 (Why is no real title available?)
- Learning sparsely used overcomplete dictionaries via alternating minimization
- Living on the edge: phase transitions in convex programs with random data
- Low-rank matrix completion using alternating minimization
- Low-rank matrix recovery with composite optimization: good conditioning and rapid convergence
- Max-Affine Regression: Parameter Estimation for Gaussian Designs
- Mixtures of linear regressions
- Non-convex optimization for machine learning
- Nonconvex Optimization Meets Low-Rank Matrix Factorization: An Overview
- Phase Retrieval Using Alternating Minimization
- Phase retrieval using alternating minimization in a batch setting
- Phase retrieval via randomized Kaczmarz: theoretical guarantees
- Phase Retrieval With Random Gaussian Sensing Vectors by Alternating Projections
- Phaseless Recovery Using the Gauss–Newton Method
- Precise Error Analysis of Regularized <inline-formula> <tex-math notation="LaTeX">$M$ </tex-math> </inline-formula>-Estimators in High Dimensions
- Precise statistical analysis of classification accuracies for adversarial training
- Randomly initialized EM algorithm for two-component Gaussian mixture achieves near optimality in \(O(\sqrt{n})\) iterations
- Selected Works of David Brillinger
- Sharp Time–Data Tradeoffs for Linear Inverse Problems
- Single-Index Models in the High Signal Regime
- Singularity, misspecification and the convergence rate of EM
- Solving (most) of a set of quadratic equalities: composite optimization for robust phase retrieval
- Some inequalities for Gaussian processes and applications
- Statistical and computational guarantees for the Baum-Welch algorithm
- Statistical guarantees for the EM algorithm: from population to sample-based analysis
- Symbol Error Rate Performance of Box-Relaxation Decoders in Massive MIMO
- The convex geometry of linear inverse problems
- The distribution of the Lasso: uniform control over sparse balls and adaptive parameter tuning
- The Generalized Lasso With Non-Linear Observations
- The landscape of empirical risk for nonconvex losses
- The LASSO Risk for Gaussian Matrices
- The Noise-Sensitivity Phase Transition in Compressed Sensing
- The nonsmooth landscape of phase retrieval
Cited in
(10)- On the Global Convergence of Randomized Coordinate Gradient Descent for Nonconvex Optimization
- Noisy linear inverse problems under convex constraints: exact risk asymptotics in high dimensions
- Stochastic algorithms with geometric step decay converge linearly on sharp functions
- Alternating minimization for generalized rank-1 matrix sensing: sharp predictions from a random initialization
- High-dimensional scaling limits and fluctuations of online least-squares SGD with smooth covariance
- Nonlinear tomographic reconstruction via nonsmooth optimization
- A new central limit theorem for the augmented IPW estimator: variance inflation, cross-fit covariance and beyond
- Hitting the high-dimensional notes: an ODE for SGD learning dynamics on GLMs and multi-index models
- Asymptotic dynamics of alternating minimization for bilinear regression
- Uncertainty quantification for iterative algorithms in linear models with application to early stopping
This page was built for publication: Sharp global convergence guarantees for iterative nonconvex optimization with random data
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6046308)
