scale_c test

scale_c.py

import numpy as np
import matplotlib.pyplot as plt

from sklearn.svm import LinearSVC
from sklearn.cross_validation import ShuffleSplit
from sklearn.grid_search import GridSearchCV

from sklearn import datasets

n_samples = 100
n_features = 1000
X, y = datasets.make_classification(n_samples=n_samples, n_features=n_features,
        n_informative=5, random_state=1)

for train_fraction in np.arange(0.1, 0.6, 0.1):
    svm = LinearSVC(penalty='L1', dual=False, scale_C=True)
    cs = 2. ** np.arange(-15, 15)

    param_grid = dict(C=cs)

    grid = GridSearchCV(svm, param_grid=param_grid,
            cv=ShuffleSplit(n=n_samples, train_fraction=train_fraction,
                n_iterations=100, random_state=1))

    grid.fit(X, y)
    scores = [x[1] for x in grid.grid_scores_]
    plt.plot(np.arange(-15, 15), scores, label="fraction %.2f" %
            train_fraction)
    print(svm)
plt.legend(loc="best")
plt.show()