Skip to main content
U.S. Department of Energy
Office of Scientific and Technical Information

Revisiting the ODE Method for Recursive Algorithms: Fast Convergence Using Quasi Stochastic Approximation

Journal Article · · Journal of Systems Science and Complexity

Several decades ago, Profs. Sean Meyn and Lei Guo were postdoctoral fellows at ANU, where they shared interest in recursive algorithms. It seems fitting to celebrate Lei Guo's 60th birthday with a review of the ODE Method and its recent evolution. The method has been regarded as a technique for algorithm analysis. It is argued that this viewpoint is backwards: The original stochastic approximation method was surely motivated by an ODE, and tools for analysis came much later (based on establishing robustness of Euler approximations). The paper presents a brief survey of recent research in machine learning that shows the power of algorithm design in continuous time, following by careful approximation to obtain a practical recursive algorithm. While these methods are usually presented in a stochastic setting, this is not a prerequisite. In fact, recent theory shows that rates of convergence can be dramatically accelerated by applying techniques inspired by quasi Monte-Carlo. Subject to conditions, the optimal rate of convergence can be obtained by applying the averaging technique of Polyak and Ruppert. The conditions are not universal, but theory suggests alternatives to achieve acceleration. The theory is illustrated with applications to gradient-free optimization, and policy gradient algorithms for reinforcement learning.

Research Organization:
National Renewable Energy Laboratory (NREL), Golden, CO (United States)
Sponsoring Organization:
USDOE National Renewable Energy Laboratory (NREL), Laboratory Directed Research and Development (LDRD) Program
DOE Contract Number:
AC36-08GO28308
OSTI ID:
1866774
Report Number(s):
NREL/JA-5D00-82852; MainId:83625; UUID:7b1f0e19-78b1-4858-b2db-e2141be91986; MainAdminID:64465
Journal Information:
Journal of Systems Science and Complexity, Vol. 34
Country of Publication:
United States
Language:
English

References (31)

Convergence rate of least-squares identification and adaptive control for stochastic systems† journal November 1986
Estimating time-varying parameters by the Kalman filter based algorithm: stability and convergence journal January 1990
A new approach to stochastic adaptive control journal March 1987
Adaptive control for time-varying systems: A combination of Martingale and Markov chain techniques journal March 1989
Stability of Recursive Stochastic Tracking Algorithms journal September 1994
A Stochastic Approximation Method journal September 1951
On the Convergence of Stochastic Iterative Dynamic Programming Algorithms journal November 1994
Analysis of recursive stochastic algorithms journal August 1977
Adaptive and Robust Control in the USSR journal January 2020
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 journal February 2000
Continuous-time stochastic Mirror Descent on a network: Variance reduction, consensus, convergence conference December 2012
A variational perspective on accelerated methods in optimization journal November 2016
On Matrix Momentum Stochastic Approximation and Applications to Q-learning conference September 2019
Sequences with low discrepancy generalisation and application to bobbins-monbo algorithm journal January 1990
Stochastic approximation with averaging innovation applied to Finance journal January 2012
Accelerating Optimization and Reinforcement Learning with Quasi Stochastic Approximation conference May 2021
Quasi-Stochastic Approximation and Off-Policy Reinforcement Learning conference December 2019
The O.D.E. Method for Convergence of Stochastic Approximation and Reinforcement Learning journal January 2000
Stability of Stochastic Approximations With “Controlled Markov” Noise and Temporal Difference Learning journal June 2019
Acceleration of Stochastic Approximation by Averaging journal July 1992
Real-Time Optimization by Extremum-Seeking Control book September 2003
Stochastic Estimation of the Maximum of a Regression Function journal September 1952
Random-seeking methods for the stochastic unconstrained optimization journal March 1975
Multivariate stochastic approximation using a simultaneous perturbation gradient approximation journal March 1992
A Kiefer-Wolfowitz algorithm with randomized differences journal March 1999
Random Gradient-Free Minimization of Convex Functions journal November 2015
Two-timescale simultaneous perturbation stochastic approximation using deterministic perturbation sequences journal April 2003
Estimation and control in Markov chains journal March 1974
Adaptive control of Markov chains, I: Finite parameter set journal December 1979
Identification and Adaptive Control of Markov Chains journal July 1982
Théorèmes de convergence presque sure pour une classe d'algorithmes stochastiques à pas décroissant journal September 1987