Heuristic Policy Analysis and Efficiency Assessment in Constraint Satisfaction Search

Richard Wallace
2006 Proceedings - International Conference on Tools with Artificial Intelligence, TAI  
This paper argues that assessments of search effort made within a new framework for characterizing heuristic performance based on adherence to optimal policies can elucidate many differences in search effort that occur from using different variable ordering heuristics. After providing a brief overview of this framework and discussing the manner in which adherence to a policy can be measured, the paper presents results from a series of experiments to show that many complex patterns of
more » ... can indeed be explained by assessments based on such measures of adherence. In particular, differences found for different categories of heuristics and differences in heuristic performance related to amount of propagation (specifically, MAC versus forward checking) can be elucidated by this approach to performance assessment.
doi:10.1109/ictai.2006.62 dblp:conf/ictai/Wallace06 fatcat:4tj76njdljf2ldfwkeptotr2dy