Solving Sudoku with genetic operations that preserve building blocks

Genetic operations that consider effective building blocks are proposed for using genetic algorithms to solve Sudoku puzzles. A stronger local search function is also proposed. Evaluation of the proposed techniques using commercial Sudoku puzzle sets and three puzzles ranked as super difficult compared with previously reported examples show that the rate of optimum solutions can be greatly improved. It is demonstrated that even further improvement in accuracy is expected from a correction technique based on disparity hypothesis.