[numbers] Genetic algorithms?

Has anyone managed to apply a genetic algorithm to this contest? I haven't seen anybody mention the non-brute-force path as of yet.

This is not an appropriate problem for genetic algorithms, I think... Because of the exactness of solutions, you can't estimate whether you're close to a solution or not, which is basically what you need to pick out "best" members of a population

I'm glad to hear you say that... I hadn't figured out how to use genetic algorithms myself, although some of the solvers out there claim to for the Countdown Numbers Game problem.

We proposed this competition to Matt, and it has been solved in the past with genetic algorithms. However it's not appropriate for finding all possible solutions.

The competition here in fact deviates slightly from Countdown because the Countdown rules have several possible large numbers (25, 50, 75, 100) and the rest are small. Numbers are on cards that are selected from a table. Typical selections are 1 large and the rest small.

A rule, which I don't personally agree with, is that using a subset of numbers is permissible when contestants search in 30 seconds for a solution.

So if the numbers were 25, 1, 6, 7, 2, 4, with a target of 300, then 25 * 6 would be an acceptable solution.

Nick / ionCube