Comparing various online solvers

An example showing how different online solvers perform on the hand-written digits dataset.

Python source code: plot_sgd_comparison.py

# Author: Rob Zinkov <rob at zinkov dot com>
# License: BSD 3 clause

import numpy as np
import pylab as pl
from sklearn import datasets
from sklearn.cross_validation import train_test_split
from sklearn.linear_model import SGDClassifier, Perceptron
from sklearn.linear_model import PassiveAggressiveClassifier

heldout = [0.95, 0.90, 0.75, 0.50, 0.01]
rounds = 20
digits = datasets.load_digits()

classifiers = [
    ("SGD", SGDClassifier()),
    ("Perceptron", Perceptron()),
    ("Passive-Aggressive I", PassiveAggressiveClassifier(loss='hinge',
                                                         C=1.0)),
    ("Passive-Aggressive II", PassiveAggressiveClassifier(loss='squared_hinge',
                                                          C=1.0)),
]

xx = 1 - np.array(heldout)
for name, clf in classifiers:
    yy = []
    for i in heldout:
        yy_ = []
        for r in range(rounds):
            X_train, X_test, y_train, y_test = train_test_split(digits.data,
                                                                digits.target,
                                                                test_size=i)
            clf.fit(X_train, y_train)
            y_pred = clf.predict(X_test)
            yy_.append(1 - np.mean(y_pred == y_test))
        yy.append(np.mean(yy_))
    pl.plot(xx, yy, label=name)

pl.legend(loc="upper right")
pl.xlabel("Proportion train")
pl.ylabel("Test Error Rate")
pl.show()