Skip to content

On how to give integers an extra bonus in score while searching #376

Answered by MilesCranmer
Eachen-Soong asked this question in Q&A

You must be logged in to vote

Hi @NoRatRacing,

Indeed concatenating constant columns is by far the easiest way to tackle this specification. Here are your options:

  1. You could create one feature column for every integer (and constant) you are interested in including. Then, you could add complexity_of_constants=5 to make regular real constants (like 10.996) more expensive – so the search will prefer your provided integers!

    I've done this approach even for integers 1 to 30, and it seems to do okay. You might need to run a bit longer though.

    Indeed it's not the most elegant though.

  2. The proper way to do this would be to introduce mixed integer programming. If someone has time to try that (with JuMP.jl), and push it, I (…

Replies: 1 comment

You must be logged in to vote
0 replies
Answer selected by Eachen-Soong
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Category
Q&A
Labels
PySR PySR-related discussion SymbolicRegression.jl SymbolicRegression.jl-related discussion
2 participants