Embedding knowledge into stochastic learning automata for fast solution of binary constraint satisfaction problems

摘要

We present & model for solving binary constraint satisfaction problems (CSPs) based on stochastic learning automata and an underlying network whose connection pattern represents the CSP. The operation of the learning automata yields fast convergence due to the incorporation of local and global knowledge about the state of the network. The model, referred to as Enhanced Stochastic Automata (ESA) model converges with probability 1 to a solution of the CSP and proves remarkably fast in dealing iwth large problems.