def test_svm_online(sd, scr):
svm = svmc.SVM()
svm.regularizer = 1.0 # This is lambda
params_svm_sd = svm.train(sd.train_X,sd.train_y)
y_pred_train = svm.test(sd.train_X,params_svm_sd)
acc_train = svm.evaluate(sd.train_y, y_pred_train)
assert allclose(acc_train, 0.940000, tolerance)
y_pred_test = svm.test(sd.test_X,params_svm_sd)
acc_test = svm.evaluate(sd.test_y, y_pred_test)
assert allclose(acc_test, 0.960000, tolerance)
params_svm_sc = svm.train(scr.train_X,scr.train_y)
y_pred_train = svm.test(scr.train_X,params_svm_sc)
acc_train = svm.evaluate(scr.train_y, y_pred_train)
assert allclose(acc_train, 0.87875, 0.01)
y_pred_test = svm.test(scr.test_X,params_svm_sc)
acc_test = svm.evaluate(scr.test_y, y_pred_test)
# TODO: py2 gives 0.805, check the reason for the different value
assert allclose(acc_test, 0.810000, 0.01)
def run_all_classifiers(dataset):
fig, axis = dataset.plot_data()
print "Naive Bayes"
nb = nbc.NaiveBayes()
params_nb = nb.train(dataset.train_X, dataset.train_y)
print params_nb.reshape(-1)
predict = nb.test(dataset.train_X, params_nb)
evaluation = nb.evaluate(predict, dataset.train_y)
predict2 = nb.test(dataset.test_X, params_nb)
evaluation2 = nb.evaluate(predict2, dataset.test_y)
print "Accuracy train: %f test: %f" % (evaluation, evaluation2)
fig, axis = dataset.add_line(fig, axis, params_nb, "Naive Bayes", "red")
print "Perceptron"
perc = percc.Perceptron()
params_perc = perc.train(dataset.train_X, dataset.train_y)
print params_perc.reshape(-1)
predict = perc.test(dataset.train_X, params_perc)
evaluation = perc.evaluate(predict, dataset.train_y)
predict2 = perc.test(dataset.test_X, params_perc)
evaluation2 = perc.evaluate(predict2, dataset.test_y)
print "Accuracy train: %f test: %f" % (evaluation, evaluation2)
fig, axis = dataset.add_line(fig, axis, params_perc, "Perceptron", "blue")
print "MaxEnt LBFGS"
me = mec_batch.MaxEnt_batch()
params_me = me.train(dataset.train_X, dataset.train_y)
print params_me.reshape(-1)
predict = me.test(dataset.train_X, params_me)
evaluation = me.evaluate(predict, dataset.train_y)
predict2 = me.test(dataset.test_X, params_me)
evaluation2 = me.evaluate(predict2, dataset.test_y)
print "Accuracy train: %f test: %f" % (evaluation, evaluation2)
fig, axis = dataset.add_line(fig, axis, params_me, "ME-LBFGS", "green")
print "MaxEnt Online"
me_online = mec_online.MaxEnt_online()
params_me = me_online.train(dataset.train_X, dataset.train_y)
print params_me.reshape(-1)
predict = me_online.test(dataset.train_X, params_me)
evaluation = me_online.evaluate(predict, dataset.train_y)
predict2 = me_online.test(dataset.test_X, params_me)
evaluation2 = me.evaluate(predict2, dataset.test_y)
print "Accuracy train: %f test: %f" % (evaluation, evaluation2)
fig, axis = dataset.add_line(fig, axis, params_me, "ME-Online", "pink")
print "MIRA"
mira = mirac.Mira()
params_mira = mira.train(dataset.train_X, dataset.train_y)
print params_mira.reshape(-1)
predict = mira.test(dataset.train_X, params_mira)
evaluation = mira.evaluate(predict, dataset.train_y)
predict2 = mira.test(dataset.test_X, params_mira)
evaluation2 = mira.evaluate(predict2, dataset.test_y)
print "Accuracy train: %f test: %f" % (evaluation, evaluation2)
fig, axis = dataset.add_line(fig, axis, params_mira, "Mira", "orange")
print "SVM"
svm = svmc.SVM()
params_svm = svm.train(dataset.train_X, dataset.train_y)
print params_svm.reshape(-1)
predict = svm.test(dataset.train_X, params_svm)
evaluation = svm.evaluate(predict, dataset.train_y)
predict2 = svm.test(dataset.test_X, params_svm)
evaluation2 = svm.evaluate(predict2, dataset.test_y)
print "Accuracy train: %f test: %f" % (evaluation, evaluation2)
fig, axis = dataset.add_line(fig, axis, params_svm, "SVM", "brown")
print
# Same as above, but for the Maximum Entropy classifier, online version
me_sgd = meoc.MaxEntOnline()
params_meo_sd = me_sgd.train(sd.train_X, sd.train_y)
y_pred_train = me_sgd.test(sd.train_X, params_meo_sd)
acc_train = me_sgd.evaluate(sd.train_y, y_pred_train)
y_pred_test = me_sgd.test(sd.test_X, params_meo_sd)
acc_test = me_sgd.evaluate(sd.test_y, y_pred_test)
fig, axis = sd.add_line(fig, axis, params_meo_sd, "Max-Ent-Online", "magenta")
print "Max-Ent Online Simple Dataset Accuracy train: %f test: %f" % (acc_train,
acc_test)
print
# Same as above, but for the SVM classifier
svm = svmc.SVM()
params_svm_sd = svm.train(sd.train_X, sd.train_y)
y_pred_train = svm.test(sd.train_X, params_svm_sd)
acc_train = svm.evaluate(sd.train_y, y_pred_train)
y_pred_test = svm.test(sd.test_X, params_svm_sd)
acc_test = svm.evaluate(sd.test_y, y_pred_test)
fig, axis = sd.add_line(fig, axis, params_svm_sd, "SVM", "yellow")
print "SVM Online Simple Dataset Accuracy train: %f test: %f" % (acc_train,
acc_test)
print
# End of exercise 3.1 #########
# Exercise 3.2: implement Naive Bayes for multinomial data ########
# Read the book review data