TY - CHAP
T1 - Tuning local search by average-reward reinforcement learning
AU - Prestwich, Steven
PY - 2008
Y1 - 2008
N2 - Reinforcement Learning and local search have been combined in a variety of ways, in order to learn how to solve combinatorial problems more efficiently. Most approaches optimise the total reward, where the reward at each action is the change in objective function. We argue that it is more appropriate to optimise the average reward. We use R-learning to dynamically tune noise in standard SAT local search algorithms on single instances. Experiments show that noise can be successfully automated in this way.
AB - Reinforcement Learning and local search have been combined in a variety of ways, in order to learn how to solve combinatorial problems more efficiently. Most approaches optimise the total reward, where the reward at each action is the change in objective function. We argue that it is more appropriate to optimise the average reward. We use R-learning to dynamically tune noise in standard SAT local search algorithms on single instances. Experiments show that noise can be successfully automated in this way.
UR - https://www.scopus.com/pages/publications/58349093519
U2 - 10.1007/978-3-540-92695-5_15
DO - 10.1007/978-3-540-92695-5_15
M3 - Chapter
AN - SCOPUS:58349093519
SN - 3540926941
SN - 9783540926948
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 192
EP - 205
BT - Learning and Intelligent Optimization - Second International Conference, LION 2007 II, Selected Papers
T2 - 2nd International Conference on Learning and Intelligent Optimization, LION 2007 II
Y2 - 8 December 2007 through 12 December 2007
ER -