Sudoku

A demo on how to solve a Sudoku puzzle with D-Wave Ocean.

Usage

python sudoku.py <sudoku file path>

For example,

python sudoku.py problem.txt

Code Overview

The idea is to describe the Sudoku puzzle as a set of constraints that our solution needs to satisfy (i.e. we are posing the puzzle as a constraint satisfaction problem). By laying down these constraints, we can get our solver to optimize over them and hopefully return a solution that satisfies all our constraints.

There are several constraints in Sudoku:

• Each cell in the Sudoku array must contain one digit
• No row may have duplicate digits
• No column may have duplicate digits
• No sub-square may have duplicate digits

Code Specifics

Input

The code takes as its input a text file containing a Sudoku puzzle in the following format:

• Rows of the puzzle are represented as a sequence of lines, one per row
• Cells in each row of the puzzle are represented as space-separated integers
• Empty cells are represented by zeros
• The file should not contain any additional lines (e.g. headers) or comments

For example,

8 2 0 9 1 0 0 0 7
9 0 0 7 0 6 8 1 2
0 1 7 8 0 0 0 9 0
0 8 0 0 0 0 9 7 0
0 5 2 0 9 3 1 8 0
6 0 0 1 8 7 0 0 0
0 7 8 0 0 9 0 5 0
3 0 0 2 5 0 7 6 0
5 0 9 3 0 1 2 0 8

Comments on the variables

• A Sudoku puzzle with n by n cells requires that each row, column, and sub-square have n unique values. Since the sub-square is a square matrix with n items, it means that n must be a square number (i.e. for a sub-square of size m by m, m * m = n). Hence in the code, the variables n and m represent:

  n == number of rows == number of columns
  m == sqrt(n) == number of sub-square rows == number of sub-square columns
  

Comments on the solver

• We are using a hybrid solver called Kerberos (specifically, hybrid.reference.KerberosSampler). It is a hybrid solver because it combines classical and quantum resources together
• We are using Kerberos because it can break down our problem into smaller chunks that could then be solved by our quantum computer. The quantum and classical solutions are then combined together, resulting in our final solution

License

Released under the Apache License 2.0. See LICENSE file.