A Warm Restart Strategy for Solving Sudoku by Sparse Optimization Methods

Abstract

This paper is concerned with the popular Sudoku problem. We proposed a warm restart strategy for solving Sudoku puzzles, based on the sparse optimization technique. Furthermore, we defined a new difficulty level for Sudoku puzzles. The efficiency of the proposed method is tested using a dataset of Sudoku puzzles, and the numerical results show that the accurate recovery rate can be enhanced from 84%+ to 99%+ using the L1 sparse optimization method.