Parameter estimation using grid search with a nested cross-validationΒΆ
The classifier is optimized by “nested” cross-validation using the GridSearchCV object.
The performance of the selected parameters is evaluated using cross-validation (different than the nested cross-validation that is used to select the best classifier).
Python source code: grid_search_digits.py
print __doc__
from pprint import pprint
import numpy as np
from scikits.learn import datasets
from scikits.learn.cross_val import StratifiedKFold
from scikits.learn.grid_search import GridSearchCV
from scikits.learn.metrics import classification_report
from scikits.learn.metrics import precision_score
from scikits.learn.metrics import recall_score
from scikits.learn.svm import SVC
################################################################################
# Loading the Digits dataset
digits = datasets.load_digits()
# To apply an classifier on this data, we need to flatten the image, to
# turn the data in a (samples, feature) matrix:
n_samples = len(digits.images)
X = digits.images.reshape((n_samples, -1))
y = digits.target
# split the dataset in two equal part respecting label proportions
train, test = iter(StratifiedKFold(y, 2)).next()
################################################################################
# Set the parameters by cross-validation
tuned_parameters = [{'kernel': ['rbf'], 'gamma': [1e-3, 1e-4],
'C': [1, 10, 100, 1000]},
{'kernel': ['linear'], 'C': [1, 10, 100, 1000]}]
scores = [
('precision', precision_score),
('recall', recall_score),
]
for score_name, score_func in scores:
clf = GridSearchCV(SVC(C=1), tuned_parameters, n_jobs=2,
score_func=score_func)
clf.fit(X[train], y[train], cv=StratifiedKFold(y[train], 5))
y_true, y_pred = y[test], clf.predict(X[test])
print "Classification report for the best estimator: "
print clf.best_estimator
print "Tuned for '%s' with optimal value: %0.3f" % (
score_name, score_func(y_true, y_pred))
print classification_report(y_true, y_pred)
print "Grid scores:"
pprint(clf.grid_points_scores_)
print
# Note the problem is too easy: the hyperparameter plateau is too flat and the
# output model is the same for precision and recall with ties in quality