Convergence and convergence rate of stochastic gradient search in the case of multiple and non-isolated extrema

DOI10.1016/J.SPA.2014.11.001zbMATH Open1357.62274arXiv0907.1020OpenAlexW2084687242MaRDI QIDQ2018557FDOQ2018557

Authors: Vladislav B. Tadić 

Publication date: 24 March 2015

Published in: Stochastic Processes and their Applications (Search for Journal in Brave)

Abstract: The asymptotic behavior of stochastic gradient algorithms is studied. Relying on results from differential geometry (Lojasiewicz gradient inequality), the single limit-point convergence of the algorithm iterates is demonstrated and relatively tight bounds on the convergence rate are derived. In sharp contrast to the existing asymptotic results, the new results presented here allow the objective function to have multiple and non-isolated minima. The new results also offer new insights into the asymptotic properties of several classes of recursive algorithms which are routinely used in engineering, statistics, machine learning and operations research.

Full work available at URL: https://arxiv.org/abs/0907.1020

Recommendations

• Convergence rates for the stochastic gradient descent method for non-convex objective functions
• Convergence Analysis of Stochastic Algorithms
• The convergence of stochastic gradient algorithms applied to learning in neural networks
• Asymptotic bias of stochastic gradient search
• scientific article; zbMATH DE number 1341059

zbMATH Keywords

convergence ratestochastic gradient methodsconvergence almost sure

Mathematics Subject Classification ID

Stochastic approximation (62L20) Stochastic programming (90C15) Semi-analytic sets, subanalytic sets, and generalizations (32B20)

Cites Work

• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Introduction to Stochastic Search and Optimization
• Semianalytic and subanalytic sets
• Stochastic approximation and its applications
• Nonlinear systems.
• Title not available (Why is that?)
• Applications of a Kushner and Clark lemma to general classes of stochastic algorithms
• Convergence of the Iterates of Descent Methods for Analytic Cost Functions
• On semi- and subanalytic geometry
• Title not available (Why is that?)
• Title not available (Why is that?)
• Gradient Convergence in Gradient methods with Errors
• Sur le problème de la division
• On gradients of functions definable in o-minimal structures
• On the stable equilibrium points of gradient systems
• The Łojasiewicz Inequality for Nonsmooth Subanalytic Functions with Applications to Subgradient Dynamical Systems
• Title not available (Why is that?)
• Estimation and annealing for Gibbsian fields
• The O.D.E. Method for Convergence of Stochastic Approximation and Reinforcement Learning
• Title not available (Why is that?)
• Maximum likelihood estimation for spatial models by Markov chain Monte Carlo stochastic approximation
• Criterion algorithms of stochastic optimization
• Analyticity, Convergence, and Convergence Rate of Recursive Maximum-Likelihood Estimation in Hidden Markov Models
• On the Almost Sure Rate of Convergence of Linear Stochastic Approximation Algorithms

Cited In (12)

• Title not available (Why is that?)
• Title not available (Why is that?)
• A Globally Convergent Stochastic Approximation
• Optimum parameters and nonasymptotic bounds on the rate of convergence of stochastic algorithms in criterial optimization problems
• Convergence analysis of smoothed stochastic gradient-type algorithm
• Infinite WARM graphs III: strong reinforcement regime
• Urns with multiple drawings and graph-based interaction
• A numerical approach to optimal coherent quantum LQG controller design using gradient descent
• On the existence of minimizers in shallow residual ReLU neural network optimization landscapes
• Two Time-Scale Stochastic Approximation with Controlled Markov Noise and Off-Policy Temporal-Difference Learning
• Almost sure convergence of stochastic gradient processes with matrix step sizes
• Online inference with multi-modal likelihood functions

Uses Software

• ICALAB

This page was built for publication: Convergence and convergence rate of stochastic gradient search in the case of multiple and non-isolated extrema

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2018557)