DocumentCode :
1454174
Title :
Dynamic programming and the graphical representation of error-correcting codes
Author :
Geman, Stuart ; Kochanek, Kevin
Author_Institution :
Div. of Appl. Math., Brown Univ., Providence, RI, USA
Volume :
47
Issue :
2
fYear :
2001
fDate :
2/1/2001 12:00:00 AM
Firstpage :
549
Lastpage :
568
Abstract :
Graphical representations of codes facilitate the design of computationally efficient decoding algorithms. This is an example of a general connection between dependency graphs, as arise in the representations of Markov random fields, and the dynamic programming principle. We concentrate on two computational tasks: finding the maximum-likelihood codeword and finding its posterior probability, given a signal received through a noisy channel. These two computations lend themselves to a particularly elegant version of dynamic programming, whereby the decoding complexity is particularly transparent. We explore some codes and some graphical representations designed specifically to facilitate computation. We further explore a coarse-to-fine version of dynamic programming that can produce an exact maximum-likelihood decoding many orders of magnitude faster than ordinary dynamic programming
Keywords :
Markov processes; computational complexity; dynamic programming; error correction codes; graph theory; maximum likelihood decoding; random processes; Markov random fields; computationally efficient decoding algorithms; decoding complexity; dependency graphs; dynamic programming; error-correcting codes; exact maximum-likelihood decoding; graphical representations; maximum-likelihood codeword; noisy channel; posterior probability; Algorithm design and analysis; Dynamic programming; Error correction codes; Iterative algorithms; Mathematics; Maximum likelihood decoding; Maximum likelihood estimation; Military computing; Probability; Random variables;
fLanguage :
English
Journal_Title :
Information Theory, IEEE Transactions on
Publisher :
ieee
ISSN :
0018-9448
Type :
jour
DOI :
10.1109/18.910574
Filename :
910574
Link To Document :
بازگشت