- From: Jeremy Wong 黃泓量 <jeremy@miko.hk>
- Date: Tue, 10 Jan 2006 10:23:46 +0800
- To: "Danny Ayers" <danny.ayers@gmail.com>, Jeremy Wong 黃泓量 <jeremy@1980.hk>
- Cc: "Semantic Web" <semantic-web@w3.org>
Not all sudoku puzzles can be solved by solely inference. I expect that any OWL implementation is an inference engine. It means that if a sudoku puzzle can be solved by 100% logical inference, then OWL engine can solve the puzzle as well. Any sudoku puzzle having unique solution could then be solved by OWL engine. The ability to solve different levels of sudoku puzzles can be an indicator of the intelligence of OWL implementation. From another view sudoku can be a mean to improve the intelligence of OWL engine. The techniques used in solving sudoku puzzles can apply on OWL implementation. In addition to simple scanning the cardinality constraints, there are multiple contingency and force chaining. They are advanced sudoku techniques that OWL implementation can include. Using rules is more flexible than relying on OWL engine, at least you can enter new rules any time you like. However, rules is not computing efficient sometimes. It is better to have OWL engine performs the inference anyway. BTW, solving most sudoku puzzles should not take more than tenth second with today's computer. You might go to a wrong direction if the engine ran as long as 5 hours. Even if it used brute-force, dancing-link algorithm already greatly reduced the time for the first solution. Anyway,.. I don't have time to implement the above concepts in OWL, either. Jeremy Wong 黃泓量 -- Miko-Kings Instruments Ltd. Web: http://www.miko.com.hk/ ----- Original Message ----- From: "Danny Ayers" <danny.ayers@gmail.com> To: "Jeremy Wong 黃泓量" <jeremy@1980.hk> Cc: "Semantic Web" <semantic-web@w3.org> Sent: Tuesday, January 10, 2006 12:54 AM Subject: Re: Solving Sudoku with OWL > On 1/9/06, Jeremy Wong 黃泓量 <jeremy@1980.hk> wrote:>> I attempt to model > the popular game Sudoku in OWL. Could someone help me> to see whether this > work is correct or not. > I had a go a little while back, but not had time to go further. I'mstill > not really sure whether they can be solved using RDFS/OWLinference alone > or whether rules would be needed (I suspect thelatter). > Code's at:http://pragmatron.org/trac/browser/pragmatron/sudoku/ > The files called sudoku*, lots of failed attempts! > Notes (slightly older) > at:http://dannyayers.com/?s=Sudoku&submit=Search+Archives > > Cheers,Danny. > -- > http://dannyayers.com >
Received on Tuesday, 10 January 2006 02:25:14 UTC