Solving Sudoku puzzles with node based Coincidence algorithm

Abstract

In Evolutionary computation, Sudoku puzzles are categorized as hard combinatorial problems. It is almost impossible to solve these puzzles using only native operations of genetic algorithms. This article presents an application of Coincidence algorithm, which is an Estimation of distribution algorithms in the class of evolutionary computation that can outperform traditional algorithms on several combinatorial problems. It makes use of both positive and negative knowledge for solving problems. The proposed method is compared with the current best known method. It significantly outperforms problem-specific GA to solve easy, medium, and hard level of Sudoku puzzles.