hw1_q2_1.py

from hw1_nnet import NeuralNet
from keras.datasets.mnist import load_data

(X_train, y_train), (X_test, y_test) = load_data()
X_train = X_train.reshape((X_train.shape[0], -1)) / 255.0
X_test = X_test.reshape((X_test.shape[0], -1)) / 255.0


nn_params = {
    'reg': 0.01,
    'seed': 1234,
    'n_iters': 50,
    'hidden_dims': [256, 128],
    'learning_rate': 0.01,
    'activation': 'relu',
    'filename': 'learning_rate: 0.01'
}

learning_rate_options = [0.01, 0.05, 0.1]
for learning_rate in learning_rate_options:
	param = {
		'learning_rate': learning_rate,
		'filename': 'learning_rate: {}'.format(learning_rate)
	}
	nn = NeuralNet(**nn_params)
	nn.set_params(**param)
	nn.fit(X_train, y_train)