Model Selection
Most models contain hyperparameters: parameters that are specified in the
constructor, and not learned from the data. ScikitLearn.jl provides
GridSearchCV to find the best set of hyper-parameter:
using ScikitLearn.GridSearch: GridSearchCV
gridsearch = GridSearchCV(LogisticRegression(), Dict(:C => 0.1:0.1:2.0))
fit!(gridsearch, X, y)
println("Best hyper-parameters: $(gridsearch.best_params_)")
See ?GridSearchCV and the scikit-learn docs for details.
Examples
- Quick start guide
- Pipelining: chaining a PCA and a logistic regression
- Concatenating multiple feature extraction methods
Random Search
RandomizedSearchCV will sample from each parameter independently.
Documentation here.
Note: The distributions have to be specified using scipy.stats (see example
below), but we hope to support Distributions.jl in the future. File an issue if
this is a pain point.