Learning Techniques
Khaled Ghédira, Bernard Dubuisson · 2013
The real learning mechanism for constraint satisfaction problems (CSPs) implicitly or explicitly manipulates specific induced constraints, whose generation, recording and reuse do not change the set of CSP solutions and in particular prevent the search from visiting futile regions. This chapter begins by defining the notions relating to the “nogood” concept. Subsequently, the chapter presents the nogood-recording (NR) algorithm, the NR-forward-checking algorithm, and the weak-commitment-NR (WC-NR) algorithm.