Closed jolars closed 2 years ago
Maybe it makes more sense to put this function into the benchopt benchmark instead. What do you think?
Yes, I think this makes more sense (replying to myself). I'll close this one and open a PR on the benchmark repo instead.
This PR adds a function
preprocess
that preprocess X and y. Zero variance features are removed from X. If X is sparse it is then we scale usingMaxAbsScaler
. If it is dense we scale usingStandardScaler
. y is centered and scaled by it's squared norm.I'm not sure if we should handle y like this. Let me know what you think.
Maybe it makes more sense to put this function into the benchopt benchmark instead. What do you think?