Output details
11 - Computer Science and Informatics
University College London
Return to search
Output 0 of 0 in the submission
Article title
Lagrange dual decomposition for finite horizon Markov decision processes
Type
D - Journal article
Title of journal
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Article number
-
Volume number
6911 LNAI
Issue number
PART 1
First page of article
487
ISSN of journal
0302-9743
Year of publication
2011
URL
-
Number of additional authors
1
Additional information
<12>Solving finite-horizon Markov Decision Processes is a known computationally difficult problem. This paper is the first to develop an algorithm based on Lagrange Duality for solving MDPs. The method was described by a reviewer as an important new algorithm for this long-standing problem. The algorithm has remarkable empirical performance, far in excess of previous approaches. We believe that this approach will become one of the standard textbook approaches to solving this class of MDPs and part of the central core of algorithms in computer science. This resulted in an invited talk at the Royal Statistical Society and Imperial College London.
Interdisciplinary
-
Cross-referral requested
-
Research group
None
Citation count
2
Proposed double-weighted
No
Double-weighted statement
-
Reserve for a double-weighted output
No
Non-English
No
English abstract
-
