SchlossLab/mikropml

Edit hyperparameter tuning vignette

Closed this issue · 3 comments

NLesniak commented
  • Add definition of hyperparameter to intro paragraph (lines 18-19)
  • When describing alpha can range between param 0 and 1, unclear what the following "inclusive" means (line 74)
  • It would be helpful to explicitly state how the model picks final hyperparameter value and use that to talk about how we know what is best for multiple splits (line 114)
  • In SVM hyperparameters - could you more explain what sigma is, "defines how far the influence of a single training example reaches" is not clear? (line 208)
  • In xgboost, hyperparameters are not defined (line 219) and confusing wording "need to be appropriate in relation" (line 220)
BTopcuoglu commented

I'll edit to have one sentence explanations for each hyperparameter but then refer people to ML courses/papers for more detail like @kelly-sovacool suggested. Do you think that's reasonable @zenalapp?

  • 👍1
zenalapp commented

Sounds good to me!

BTopcuoglu commented

I'm done with my edits. We can close this once merge #232