steve-v / weasel Goto Github PK

Investigate whether using getDistance() for the score function will result in faster convergence than the current binary (+1 for wrong +0 for correct) score.

It's already mostly implemented, but commented out.

Use argc/argv to check the command line for a mutation/breeding switch and then use that method.

Investigate whether allowing children to breed will result in faster convergence to target.

We could achieve this by taking the best-scoring n% of children (perhaps 10%?) and breeding them together.
Currently we take a string and for each element in the string we either keep it or replace it with a random character (mutation).
Perhaps we could take two strings and build a new child string by randomly selecting characters from each of the two parents?