def main():
print('Loading data...')
mnist = np.loadtxt('data/mnist_test.csv', delimiter=',')
X_train = mnist[:len(mnist)//2, 1:]
y_train = mnist[:len(mnist)//2, 0].astype(np.int)
X_test = mnist[len(mnist)//2:, 1:]
y_test = mnist[len(mnist)//2:, 0].astype(np.int)
print('Training Sklearn OVR...')
y_pred_train, y_pred_test = sklearn_multiclass_prediction(
'ovr', X_train, y_train, X_test)
print('Sklearn OVR Accuracy (train):',
metrics.accuracy_score(y_train, y_pred_train))
print('Sklearn OVR Accuracy (test) :',
metrics.accuracy_score(y_test, y_pred_test))
print('Training Sklearn OVO...')
y_pred_train, y_pred_test = sklearn_multiclass_prediction(
'ovo', X_train, y_train, X_test)
print('Sklearn OVO Accuracy (train):',
metrics.accuracy_score(y_train, y_pred_train))
print('Sklearn OVO Accuracy (test) :',
metrics.accuracy_score(y_test, y_pred_test))
print('Training Sklearn Crammer-Singer...')
y_pred_train, y_pred_test = sklearn_multiclass_prediction(
'crammer', X_train, y_train, X_test)
print('Sklearn Crammer-Singer Accuracy (train):',
metrics.accuracy_score(y_train, y_pred_train))
print('Sklearn Crammer-Singer Accuracy (test) :',
metrics.accuracy_score(y_test, y_pred_test))
print('Training self OVR...')
self_ovr = MulticlassSVM('ovr')
self_ovr.fit(X_train, y_train)
print('Self OVR Accuracy (train):',
metrics.accuracy_score(y_train, self_ovr.predict(X_train)))
print('Self OVR Accuracy (test) :',
metrics.accuracy_score(y_test, self_ovr.predict(X_test)))
print('Training self OVO...')
self_ovo = MulticlassSVM('ovo')
self_ovo.fit(X_train, y_train)
print('Self OVO Accuracy (train):',
metrics.accuracy_score(y_train, self_ovo.predict(X_train)))
print('Self OVO Accuracy (test) :',
metrics.accuracy_score(y_test, self_ovo.predict(X_test)))
print('Training self Crammer-Singer...')
self_cs = MulticlassSVM('crammer-singer')
self_cs.fit(X_train, y_train)
print('Self Crammer-Singer Accuracy (train):',
metrics.accuracy_score(y_train, self_cs.predict(X_train)))
print('Self Crammer-Singer Accuracy (test) :',
metrics.accuracy_score(y_test, self_cs.predict(X_test)))
def test_sklearn_crammer_accuracy(self):
y_pred_train, y_pred_test = sklearn_multiclass_prediction(
'crammer', self.X_train, self.y_train, self.X_test)
train_acc = metrics.accuracy_score(self.y_train, y_pred_train)
test_acc = metrics.accuracy_score(self.y_test, y_pred_test)
self.assertTrue(abs(train_acc - 1.0) < 5e-3)
self.assertTrue(abs(test_acc - 0.85825) < 5e-3)
def test_sklearn_ovo_accuracy(self):
y_pred_train, y_pred_test = sklearn_multiclass_prediction(
'ovo', self.X_train, self.y_train, self.X_test)
train_acc = metrics.accuracy_score(self.y_train, y_pred_train)
test_acc = metrics.accuracy_score(self.y_test, y_pred_test)
self.assertTrue(abs(train_acc - 1.0) < 1e-6)
self.assertTrue(abs(test_acc - 0.892625) < 1e-6)
def test_sklearn_ovr_accuracy(self):
y_pred_train, y_pred_test = sklearn_multiclass_prediction(
'ovr', self.X_train, self.y_train, self.X_test)
train_acc = metrics.accuracy_score(self.y_train, y_pred_train)
test_acc = metrics.accuracy_score(self.y_test, y_pred_test)
self.assertTrue(abs(train_acc - 1.0) < 1e-6)
print('test_acc = ' + str(test_acc))
self.assertTrue(abs(test_acc - 0.814875) < 1e-6)
def test_sklearn_ovr_accuracy(self):
y_pred_train, y_pred_test = sklearn_multiclass_prediction(
'ovr', self.X_train, self.y_train, self.X_test)
train_acc = metrics.accuracy_score(self.y_train, y_pred_train)
test_acc = metrics.accuracy_score(self.y_test, y_pred_test)
print("my0 train_acc", abs(train_acc - 1.0))
print("my0 test_acc", abs(test_acc - 0.814875))
self.assertTrue(abs(train_acc - 1.0) < 5e-3)
self.assertTrue(abs(test_acc - 0.814875) < 5e-3)
if __name__ == '__main__':
print('Loading data...')
mnist = np.loadtxt('data/mnist_test.csv', delimiter=',')
X_train = mnist[:len(mnist) // 2, 1:]
print(np.shape(X_train))
y_train = mnist[:len(mnist) // 2, 0].astype(np.int)
print(y_train)
X_test = mnist[len(mnist) // 2:, 1:]
y_test = mnist[len(mnist) // 2:, 0].astype(np.int)
print(np.shape(X_test))
print(np.shape(y_test))
print('Training Sklearn OVR...')
y_pred_train, y_pred_test = sklearn_multiclass_prediction(
'ovr', X_train, y_train, X_test)
print('Sklearn OVR Accuracy (train):',
metrics.accuracy_score(y_train, y_pred_train))
print('Sklearn OVR Accuracy (test) :',
metrics.accuracy_score(y_test, y_pred_test))
print("pred_test result:", y_pred_test)
print("pred test shape:", np.shape(y_pred_test))
print('Training Sklearn OVO...')
y_pred_train, y_pred_test = sklearn_multiclass_prediction(
'ovo', X_train, y_train, X_test)
print('Sklearn OVO Accuracy (train):',
metrics.accuracy_score(y_train, y_pred_train))
print('Sklearn OVO Accuracy (test) :',
metrics.accuracy_score(y_test, y_pred_test))
