Generation of stationary control policies with best expected performance for a family of Markov chains

Noah Berlow, Ranadip Pal

Research output: Contribution to journalArticlepeer-review

1 Scopus citations

Abstract

Due to the high computational cost associated with calculation of optimal control policies for a family of Markov chains, sub-optimal policy generation approaches are often required for development of intervention strategies in systems medicine. In this paper, a new infinite-horizon suboptimal policy generation algorithm using a sequential-selection approach with efficient re-calculation of steady-state distributions is presented and compared against brute-force and other sub-optimal algorithms. Our results show that for a system represented by a family of Markov chains, the presented approach is able to generate robust stationary control policies with superior expected behavior in a computationally inexpensive manner.

Original languageEnglish
Pages (from-to)423-440
Number of pages18
JournalJournal of Biological Systems
Volume20
Issue number4
DOIs
StatePublished - Dec 2012

Keywords

  • Genetic Regulatory Network Control
  • Perturbation of Markov chains
  • Stationary Control Policy

Fingerprint Dive into the research topics of 'Generation of stationary control policies with best expected performance for a family of Markov chains'. Together they form a unique fingerprint.

Cite this