Each year freshmen in my school present a science experiment to show potential future freshmen that science is cool. As everybody did Physics-only projects, I decided with a friend to do a "math experiment" and we didn't have any idea other than presenting an algorithm : each one of us will do it's implementation, (this one and a home-made one) and we'll measure them to tell people that "using Maths can speed up things and solve complex day to day problems". This is an implementation of a Sudoku solver in Java, using Knuth's DLX algorithm. This algorithm solves the NP-complete exact cover set problem in a reasonable time. The software converts your sudoku table into an exact cover set problem matrix before solving it. Take note that the hardest sudoku available is solved in almost 10 seconds on a 1.83GHz Core Duo processor. Please take care that I didn't have a lot of time to program this, so the code is CRAPPY. But the exact cover set problem can modelize other constrained situations so this implementation may be a good start for other people. NB : Note that I used Mathematica to pilot directly the classes using their Java bridge, so there isn't much of an user interface.