AUTOMATIC GENERATION OF FPTASes FOR STOCHASTIC MONOTONE DYNAMIC PROGRAMS MADE EASIER

Tzvi Alon, And N.I.R. Halman

Research output: Contribution to journalArticlepeer-review

2 Scopus citations

Abstract

In this paper we go one step further in the automatic generation of FPTASes for multistage stochastic dynamic programs with scalar state and action spaces, in which the cost-to-go functions have a monotone structure in the state variable. While there exist a few frameworks for automatic generation of FPTASes, so far none of them is general and simple enough to be extensively used. We believe that our framework has these two attributes and has great potential to attract interest from both the operations research and theoretical computer science communities. Moreover, it seems very reasonable that many intractable problems that currently do not admit an FPTAS, can be formulated as DPs that fit into our framework and therefore will admit a first FPTAS. Our results are achieved by a combination of Bellman equation formulations, the technique of K-approximation sets and functions, and in particular the calculus of K-approximation functions.

Original languageEnglish
Pages (from-to)2679-2722
Number of pages44
JournalSIAM Journal on Discrete Mathematics
Volume35
Issue number4
DOIs
StatePublished - 2021
Externally publishedYes

Bibliographical note

Publisher Copyright:
© 2021 Society for Industrial and Applied Mathematics.

Keywords

  • FPTAS
  • K-approximation sets
  • dynamic programming
  • functions

Fingerprint

Dive into the research topics of 'AUTOMATIC GENERATION OF FPTASes FOR STOCHASTIC MONOTONE DYNAMIC PROGRAMS MADE EASIER'. Together they form a unique fingerprint.

Cite this