CSP-Rules V2.1

[denis.be...@gmail.com]

This is to announce the publication of CSP-Rules-V2.1

CSP-Rules is a general pattern-based (or rule-based) solver of finite binary Constraint Satisfaction Problems.

V2.1 is the first public release.

CSP-Rules is inherently associated with the approach to CSP solving defined and largely illustrated in my book “Pattern-Based Constraint Satisfaction and Logic Puzzles” (PBCS).

A priori, CSP-Rules deals only with binary constraints, but the applications studied in [PBCS] and included in the CSP-Rules-V2.1 package (namely: solvers for Latin Squares, Sudoku, Futoshiki, Kakuro, Map Colouring, Numbrix, Hidato and Slithering) show that  many types of non-binary constraints can be efficiently transformed into binary ones by adding problem-specific CSP-Variables, thus making them amenable to the CSP-Rules treatment.

Both the oldest and the current implementations of CSP-Rules are based on CLIPS.

As I don't want to overload lots of mailboxes with a long email, I direct you to the GitHub website for more information: https://github.com/denis-berthier/CSP-Rules-V2.1

[denis.be...@gmail.com]

Creating unbiased collections of instances for some Constraint Satisfaction Problems may be extremely hard, even in highly structured problems such as Sudoku. The most obvious algorithms are computationally impractical or have a very strong and impossible to compute bias.

I've just published on GitHub (https://github.com/denis-berthier/Controlled-bias_Sudoku_generator_and_collection) the software used to create a "controlled-bias" collection of Sudoku puzzles: it is still biased (though much less than the previously know algorithms) but the main difference is, the bias is known.

The repository also contains the collection itself, the result of several months of CPU times.

As an application of the CSP-Rules software,  it also contains the full classification of this collection according to the W rating defined in the book mentioned in my previous post [PBCS], also the result of much CPU time.

Those are very old results, but I hadn't had time before for putting all the bits together.