Nesting of operators

[toniokroeger]

Hi! I have been using PYSR for a project and have had a lot of success with it - thanks for providing such a useful package!

One thing I have been noticing in my outputs is a strong preference for the regression to nest operators as few times as possible. For example, say that an equality a+b=c+d holds on parts of the data. This could also be expressed as a=c-b+d, but i find that the algorithms output this other form much more infrequently. Is there any way to modify things so that these two expressions are preferred at more similar rates?

I am aware of template expressions, but don't think this would apply in my case, since it could compromise `fairness' requirements for my learning process.

Thanks for taking the time to read - hope to hear back soon! Cheers.

[MilesCranmer]

Do you mean have c - b + d preferred over c + d? The former would be higher complexity so is ranked lower. Unless I misunderstand