Scores versus losses being addressed properly in TreeParzen?

JuliaAI / MLJTuning.jl

Hyperparameter optimization algorithms for use in the MLJ machine learning framework

MIT License

67 stars 12 forks source link

Scores versus losses being addressed properly in TreeParzen? #136

Closed ablaom closed 2 years ago

ablaom commented 3 years ago

That is, do we make sure we are getting the sign correct.

lnguyenPHI commented 3 years ago

TreeParzen indeed doesn't consider the measurement orientation when making suggestions. The fix should be as simple as multiplying by -1 for score measures. I'll submit a PR later today!

ablaom commented 3 years ago

Please copy me into the PR, thanks.

ablaom commented 2 years ago

@lhnguyen-vn We can close this, right?