Single-pass Self Parameter Tuning #1553
Draft
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This has been in the works for quite a long time, actually.
SSPT works by self-tuning the hyperparameters of models while keeping their parameters intact. For example, we might change some hyperparameters of linear models while keeping the weights learned so far intact. This is different from instantiating a new model from scratch with a set of candidate hyperparameter values.
@BrunoMVeloso is one of the authors and he is going to help me with the final touches.
The main things to do:
About the last item, the way SSPT is structured right now makes it inherit directly from
base.Estimator
. The good part is that you can use it to work with classifiers, regressors, clustering algorithms, and anomaly detectors (there is no need for aSSPTClassifier
and so on, for example). The downside is that SSPTcurrently does not work with
evaluate.progressive_val_score` and friends because the evaluators check internally whether the estimator is a classifier, regressor, and whatnot.