A one-bit, comparison-based gradient estimator

DOI10.1016/J.ACHA.2022.03.003MaRDI QIDQ2155805zbMATH OpenOpenAlexFDO

Authors HanQin Cai, Daniel McKenzie, Wotao Yin, Zhen-Liang Zhang

Publication date 15 July 2022

Published in Applied and Computational Harmonic Analysis (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/2010.02479

zbMATH Keywords

reinforcement learninghyperparameter tuningzeroth-order optimizationone-bit compressed sensingcomparison-based optimizationnormalized gradient descent

Mathematics Subject Classification ID

Computer science (68-XX) Information and communication theory, circuits (94-XX)

Recommendations

• Small errors in random zeroth-order optimization are imaginary
• Minimax efficient finite-difference stochastic gradient estimators using black-box function evaluations
• Black-box reductions for zeroth-order gradient algorithms to achieve lower query complexity
• Zeroth-order regularized optimization (ZORO): approximately sparse gradients and adaptive sampling

Cites work

• scientific article; zbMATH DE number 3910150 (Why is no real title available?)
• scientific article; zbMATH DE number 845714 (Why is no real title available?)
• scientific article; zbMATH DE number 6276119 (Why is no real title available?)
• A dimensionality reduction technique for unconstrained global optimization of functions with low effective dimensionality
• Active subspaces. Emerging ideas for dimension reduction in parameter studies
• Adaptive stochastic approximation by the simultaneous perturbation method
• Bayesian optimization in a billion dimensions via random embeddings
• CoSaMP: Iterative signal recovery from incomplete and inaccurate samples
• Derivative-free optimization methods
• Minimax number of strata for online stratified sampling given noisy samples
• Parallel distributed block coordinate descent methods based on pairwise comparison oracle
• Preference-based reinforcement learning: a formal framework and a policy iteration algorithm
• Restricted strong convexity and its applications to convergence analysis of gradient-type methods in convex optimization
• Robust 1-bit Compressed Sensing and Sparse Logistic Regression: A Convex Programming Approach
• Zeroth-order regularized optimization (ZORO): approximately sparse gradients and adaptive sampling

Cited in

(4)

• Global solutions to nonconvex problems by evolution of Hamilton-Jacobi PDEs
• Small errors in random zeroth-order optimization are imaginary
• Stochastic zeroth order descent with structured directions
• A One-bit, Comparison-Based Gradient Estimator

This page was built for publication: A one-bit, comparison-based gradient estimator

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2155805)