Frank is a bot that does science. He takes a dataset and forms hypothesises about mathematical constants (aka conservation laws) of the data. He then tests the hypothesis.
The Buckingham Pi Theorem could probably be used as a starting point in the hypothesis forming. (The wiki article about it is sort of helpful if you don't know what it is.)
Frank could guess at a couple of attributes that are important and then using the units of these attributes reduce it into only a few non-dimensional relevant parameters.
This should reduce the number of crap ideas and also incease the convergence of the non-crap ones.
The Buckingham Pi Theorem could probably be used as a starting point in the hypothesis forming. (The wiki article about it is sort of helpful if you don't know what it is.)
Frank could guess at a couple of attributes that are important and then using the units of these attributes reduce it into only a few non-dimensional relevant parameters.
This should reduce the number of crap ideas and also incease the convergence of the non-crap ones.