View Full Version : sudoku@vtaiwan

30th July 2011, 08:36 AM


sudoku@vtaiwan is one of V-Taiwan projects that uses Internet-connected computers to do research in sudoku. You can participate as a volunteer by downloading and running programs on your computer.

In this project, we develops new techniques and used them to modify the program Checker (written by Gary McGuire). We successfully reduced the total expected computation time from 300,000 years per core to 2,417 years per core. This makes it more feasible and reasonable to use BOINC to solve Sudoku. Some of these new techniques are described in a paper in the IWCG Workshop of TAAI 2010 conference.

22nd August 2011, 11:28 PM
An interesting little project.

Specifically, what they are trying to prove/disprove is whether there is any sudoku having only 16 numbers as clues that leads to a unique solution (some 50,000 + 17 clue solutions have been found). So this is a brute force approach to search all 16 clue sudoku setups and check if there is a unique solution to any of them.

So far, they have worked through about 10% of the space in about 8 months. Should the project ever become widely popular, I imagine it would be finished in relatively short order. It's a finite question.

The question then becomes ... what then? If they actually find a 16 clue unique solution, I suppose they would then try to find a 15 clue solution. My impression is that they do not expect to find a 16 clue solution.

Reading the forums, this question has been asked. The response has been "Also, beside the Sudoku, we have many other applications which might also use BOINC in the future such like Go, Connect6, Chinese Chess, Renju, and some puzzle games." So nothing specific at this point. The other question that would occur to me is, would this be the same project (collecting points under the same major project, similar to primegrid, or an entirely new project?