MACHINE LEARNING

Question

Grid Search
Look at the following code for running a grid search.

param_grid = {
'max_depth': [5, 15, 25, 35],
'max_leaf_nodes': [10, 20, 35, 50, 75]}
dt = DecisionTreeClassifier()
gs = GridSearchCV(dt, param_grid, scoring='f1', cv=5)
gs.fit(X, y)

How many different models are being compared?

Answer 1

Review Lesson 34 that explains about grid search.
When in doubt, you can also look at the sklearn documentation.
https://scikit-learn.org/stable/modules/generated/sklearn.model_selection.GridSearchCV.html

The point of grid search, is to try running the same model with different parameters, because we intuitively may not guess correctly, which parameter value results in the best predicting power.
So the parameters are passed in param_grid. This is a dictionary data structure, where the values contain a list of each possible argument. The total number of models come from the combination of all these values (Cartesian product).

So practically you just multiply the sizes of the param_grid values.

Answer 2

20