Alternating maximization procedure for finding the global maximum of directed information

Iddo Naiss, Haim H. Permuter

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Scopus citations

Abstract

We extend the Blahut-Arimoto algorithm for maximizing Massey's directed information, which can be used for estimating the capacity of channels with delayed feedback. In order to do so, we apply ideas from the regular Blahut-Arimoto algorithm, i.e., the alternating maximization procedure, onto our new problem, and show its convergence to the global optimum value. Our main insight in this paper is that in order to find the maximum of the directed information over causal conditioning probability mass function, one can use a backward index time maximization combined with the alternating maximization procedure. We give a detailed description of the algorithm, and state its complexity and memory needed. A numerical example is presented.

Original languageEnglish
Title of host publication2010 IEEE 26th Convention of Electrical and Electronics Engineers in Israel, IEEEI 2010
Pages545-549
Number of pages5
DOIs
StatePublished - 1 Dec 2010
Event2010 IEEE 26th Convention of Electrical and Electronics Engineers in Israel, IEEEI 2010 - Eilat, Israel
Duration: 17 Nov 201020 Nov 2010

Publication series

Name2010 IEEE 26th Convention of Electrical and Electronics Engineers in Israel, IEEEI 2010

Conference

Conference2010 IEEE 26th Convention of Electrical and Electronics Engineers in Israel, IEEEI 2010
Country/TerritoryIsrael
CityEilat
Period17/11/1020/11/10

Keywords

  • Alternating maximization procedure
  • Backward index time maximization
  • Blahut-Arimoto algorithm
  • Causal conditioning
  • Channels with feedback
  • Directed information
  • Finite state channels

ASJC Scopus subject areas

  • Electrical and Electronic Engineering

Fingerprint

Dive into the research topics of 'Alternating maximization procedure for finding the global maximum of directed information'. Together they form a unique fingerprint.

Cite this