def main():
num_classes = 3
X, y = get_master_data()
train_X, train_y, val_X, val_y, test_X, test_y = split_data(X, y)
# hyperparams
learning_rates = [1]
reg_strengths = [0]
num_iters = 5000
best_model = None
best_accuracy = -1
results = {}
for lr in learning_rates:
for rs in reg_strengths:
this_model = LinearClassifier(X.shape[1], num_classes)
this_model.train(train_X, train_y, lr, rs, num_iters)
y_pred = this_model.predict(val_X)
val_accuracy = np.mean(y_pred == val_y)
y_pred = this_model.predict(train_X)
train_accuracy = np.mean(y_pred == train_y)
print('This val accuracy: ' + str(val_accuracy))
if (val_accuracy > best_accuracy):
best_model = this_model
best_accuracy = val_accuracy
results[(lr, rs)] = train_accuracy, val_accuracy
this_model.print_model()
for lr, rs in sorted(results):
train_accuracy, val_accuracy = results[(lr, rs)]
print('lr %e reg %e train_accuracy %f val_accuracy %f' %
(lr, rs, train_accuracy, val_accuracy))
print(test_X[0])
y_pred_test = best_model.predict(test_X)
test_accuracy = np.mean(y_pred_test == test_y)
print('This test accuracy: ' + str(test_accuracy))
best_model.toFile('rowan_rest_grip_flex_linear_classifier_3_' +
str(int(test_accuracy * 100)) + 'p.csv')
