Approaching Quartic Convergence Rates for Quasi-Stochastic Approximation with Application to Gradient-Free Optimization

被引:0
作者
Lauand, Caio Kalil [1 ]
Meyn, Sean [1 ]
机构
[1] Univ Florida, Gainesville, FL 32611 USA
来源
ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 35 (NEURIPS 2022) | 2022年
基金
美国国家科学基金会;
关键词
RATIONAL LINEAR FORM; LOGARITHMS; ALGORITHMS;
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Stochastic approximation is a foundation for many algorithms found in machine learning and optimization. It is in general slow to converge: the mean square error vanishes as O(n(-1)). A deterministic counterpart known as quasi-stochastic approximation is a viable alternative in many applications, including gradient-free optimization and reinforcement learning. It was assumed in prior research that the optimal achievable convergence rate is O(n(-2)). It is shown in this paper that through design it is possible to obtain far faster convergence, of order O(n(-4+delta)), with delta > 0 arbitrary. Two techniques are introduced for the first time to achieve this rate of convergence. The theory is also specialized within the context of gradient-free optimization, and tested on standard benchmarks. The main results are based on a combination of novel application of results from number theory and techniques adapted from stochastic approximation theory.
引用
收藏
页数:14
相关论文
共 50 条
[1]  
[Anonymous], 2003, Introduction to stochastic search and optimization: estimation, simulation, and control
[2]  
Asmussen S., 2007, Stochastic Modelling and Applied Probability, V57
[3]   The heavy ball with friction method, I. The continuous dynamical system: Global exploration of the local minima of a real-valued function by asymptotic analysis of a dissipative dynamical system [J].
Attouch, H ;
Goudou, X ;
Redont, P .
COMMUNICATIONS IN CONTEMPORARY MATHEMATICS, 2000, 2 (01) :1-34
[4]  
Bernstein A, 2019, IEEE DECIS CONTR P, P5244, DOI 10.1109/CDC40024.2019.9029247
[5]   Multiscale chaotic SPSA and smoothed functional algorithms for simulation optimization [J].
Bhatnagar, S ;
Borkar, VS .
SIMULATION-TRANSACTIONS OF THE SOCIETY FOR MODELING AND SIMULATION INTERNATIONAL, 2003, 79 (10) :568-580
[6]  
BHATNAGAR S, 2013, LECT NOTES CONTROL I
[7]  
Borkar V. S., 2021, Stochastic Approximation: A Dynamical Systems Viewpoint, V2nd
[8]  
Bugeaud Y, 2018, IRMA LECT MATH THEOR, V28, P1, DOI 10.4171/183
[9]   Revisiting the ODE Method for Recursive Algorithms: Fast Convergence Using Quasi Stochastic Approximation [J].
Chen Shuhang ;
Devraj, Adithya ;
Berstein, Andrey ;
Meyn, Sean .
JOURNAL OF SYSTEMS SCIENCE & COMPLEXITY, 2021, 34 (05) :1681-1702
[10]   ON A STOCHASTIC APPROXIMATION METHOD [J].
CHUNG, KL .
ANNALS OF MATHEMATICAL STATISTICS, 1954, 25 (03) :463-483