Sudoku Solver Comparison A comparative analysis of algorithms for solving Sudoku

What is a Sudoku Puzzle?  A pencil-and-paper puzzle, much like a numeric crossword puzzle –A special type of latin square –Seen in many newspapers, including our own K-State Collegian  A highly-connected CSP –Typical 9 x 9 configuration 81 variables, each constrained by 24 other variables Total of 972 constraints A valid solution is a 9-coloring of the constraint graph

Sudoku Rules  Common Sudoku puzzles are a 9 x 9 grid of 81 cells –There are 9 rows and 9 columns –Also divided into 9 3 x3 boxes –Each cell can hold one number, an integer between 1 and 9, inclusive –Some subset of the cells are given –Each number can only appear once in each row, column, and box –Valid Sudoku have enough cells given that there is a unique solution

Sudoku images Given Solved

Algorithms  General constraint satisfaction algorithms –Backtracking search A “brute force” approach Serves as the baseline –Backtracking with MRV Look for values that are the “most constrained” in the current state  Sudoku specific algorithms –Human-like approach Avoid guessing (and backtracking!) –Some additional constraints can be deduced from values of non-adjacent cells