Efficient sampling in approximate dynamic programming algorithms

被引:21
|
作者
Cervellera, Cristiano [1 ]
Muselli, Marco [1 ]
机构
[1] Ist Studi Sistemi Intelligenti Lautomaz, Consiglio Nazl Ric, I-16149 Genoa, Italy
关键词
stochastic optimal control problem; dynamic programming; sample complexity; deterministic learning; low-discrepancy sequences;
D O I
10.1007/s10589-007-9054-8
中图分类号
C93 [管理学]; O22 [运筹学];
学科分类号
070105 ; 12 ; 1201 ; 1202 ; 120202 ;
摘要
Dynamic Programming (DP) is known to be a standard optimization tool for solving Stochastic Optimal Control (SOC) problems, either over a finite or an infinite horizon of stages. Under very general assumptions, commonly employed numerical algorithms are based on approximations of the cost-to-go functions, by means of suitable parametric models built from a set of sampling points in the d-dimensional state space. Here the problem of sample complexity, i.e., how "fast" the number of points must grow with the input dimension in order to have an accurate estimate of the cost-to-go functions in typical DP approaches such as value iteration and policy iteration, is discussed. It is shown that a choice of the sampling based on low-discrepancy sequences, commonly used for efficient numerical integration, permits to achieve, under suitable hypotheses, an almost linear sample complexity, thus contributing to mitigate the curse of dimensionality of the approximate DP procedure.
引用
收藏
页码:417 / 443
页数:27
相关论文
共 50 条
  • [1] Efficient sampling in approximate dynamic programming algorithms
    Cristiano Cervellera
    Marco Muselli
    Computational Optimization and Applications, 2007, 38 : 417 - 443
  • [2] Efficient Execution of Dynamic Programming Algorithms on Apache Spark
    Javanmard, Mohammad Mahdi
    Ahmad, Zafar
    Zola, Jaroslaw
    Pouchet, Louis-Noel
    Chowdhury, Rezaul
    Harrison, Robert
    2020 IEEE INTERNATIONAL CONFERENCE ON CLUSTER COMPUTING (CLUSTER 2020), 2020, : 337 - 348
  • [3] THE IMPORTANCE OF DESIGNING EFFICIENT ALGORITHMS USING DYNAMIC PROGRAMMING
    Munoz Matute, Judit
    Alberdi Celaya, Elisabete
    10TH INTERNATIONAL CONFERENCE OF EDUCATION, RESEARCH AND INNOVATION (ICERI2017), 2017, : 3601 - 3610
  • [4] Approximate dynamic programming algorithms for multidimensional flexible production-inventory problems
    Cimen, Mustafa
    Kirkbride, Chris
    INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH, 2017, 55 (07) : 2034 - 2050
  • [5] A Survey of Approximate Dynamic Programming
    Wang Lin
    Peng Hui
    Zhu Hua-yong
    Shen Lin-cheng
    2009 INTERNATIONAL CONFERENCE ON INTELLIGENT HUMAN-MACHINE SYSTEMS AND CYBERNETICS, VOL 2, PROCEEDINGS, 2009, : 396 - 399
  • [6] Evolutionary algorithms and dynamic programming
    Doerr, Benjamin
    Eremeev, Anton
    Neumann, Frank
    Theile, Madeleine
    Thyssen, Christian
    THEORETICAL COMPUTER SCIENCE, 2011, 412 (43) : 6020 - 6035
  • [7] Time Efficient Segmented Technique for Dynamic Programming Based Algorithms with FPGA Implementation
    Bonny, Talal
    Al Debsi, Ridhwan
    Almourad, Mohamed Basel
    JOURNAL OF CIRCUITS SYSTEMS AND COMPUTERS, 2019, 28 (13)
  • [8] Random sampling of states in dynamic programming
    Atkeson, Christopher G.
    Stephens, Benjamin J.
    IEEE TRANSACTIONS ON SYSTEMS MAN AND CYBERNETICS PART B-CYBERNETICS, 2008, 38 (04): : 924 - 929
  • [9] Separable dynamic programming and approximate decomposition methods
    Bertsekas, Dimitri P.
    IEEE TRANSACTIONS ON AUTOMATIC CONTROL, 2007, 52 (05) : 911 - 916
  • [10] Approximate dynamic programming approach for process control
    Lee, Jay H.
    INTERNATIONAL CONFERENCE ON CONTROL, AUTOMATION AND SYSTEMS (ICCAS 2010), 2010, : 459 - 464