Article ID Journal Published Year Pages File Type
376955 Artificial Intelligence 2012 17 Pages PDF
Abstract

In this paper, we study the relation among the parameters in their most general setting that define a large class of random CSP models d-k-CSP where d is the domain size and k is the length of the constraint scopes. The model d-k-CSP unifies several related models such as the model RB and the model k-CSP. We prove that the model d-k-CSP exhibits exact phase transitions if increases no slower than the logarithm of the number of variables. A series of experimental studies with interesting observations are carried out to illustrate the solubility phase transition and the hardness of instances around phase transitions.

Related Topics
Physical Sciences and Engineering Computer Science Artificial Intelligence