Published Articles >> Table of Contents >> Abstract

Fourth IEEE Symposium on Bioinformatics and Bioengineering (BIBE'04)   p. 337
Markov Decision Processes Based Optimal Control Policies for Probabilistic Boolean Networks
Osman Abul, University of Calgary, Alberta, Canada; Middle East Technical University, Ankara, Turkey
Reda Alhajj, University of Calgary, Alberta, Canada
Faruk Polat, Middle East Technical University, Ankara, Turkey

Full Article Text:

DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/BIBE.2004.1317363
Send link to a friend

Abstract
This paper addresses the control formulation process for probabilistic boolean genetic networks. It is a major problem that has not been investigated enough yet. We argue that a monitoring stage is necessary after the control stage for providing guidance about the evolution of the investigated state. For this purpose, we developed methods for generating optimal control policies for each of the following five cases: finite control, infinite control, finite control-infinite monitoring, finite control-finite monitoring, and repeated finite control-finite monitoring. Our initial proposal was based on using action cost functions in the process. In this study, we propose Markov decision processes as an alternative to the action cost functions approach. We conducted experiments on two simple illustrative examples to demonstrate that the considered five cases are necessary, effective and really matter while developing optimal control policies; the obtained results are promising.
Back to Top Additional Information
Index Terms- probabilistic boolean networks, optimal control, Markov decision processes, monitoring Citation:  Osman Abul, Reda Alhajj, Faruk Polat, "Markov Decision Processes Based Optimal Control Policies for Probabilistic Boolean Networks," bibe, p. 337,  Fourth IEEE Symposium on Bioinformatics and Bioengineering (BIBE'04),  2004