A Retrievable GA for Solving Sudoku Puzzles A Retrievable GA for Solving Sudoku Puzzles

In this paper we propose a modification to Genetic Algorithm which we call as Retrievable Genetic Algorithm (Ret-GA) and use it to solve a given Sudoku puzzle. Although a number of approaches exist for solving a given Sudoku puzzle, very few papers with heuristic approaches are available for obtaining its solution. In this work, the fitness function is modeled in a new way considering puzzle-character-dependent constraints. The GA is named as Retrievable, since the population is reinitialized after a certain number of generations in order to escape the local maxima and retrieve the solution in reduced computational time. A set of 9 sample puzzles have been considered for comparison with the previously published results by GA and it is concluded that Ret-GA performs better than GA in most of the cases. In addition we consider a set of other 75 puzzles namely Easy, Medium and Hard (25 of each type) to further test the effectiveness of proposed algorithm.