P = 200
Q = 100
learning_rate = 0.05
t_max = 100 #total epochs
trainX, trainY, testX, testY = read_file("xi(1).csv", "tau(1).csv", P, Q)
Es = []
E_tests = []
for i in range(0, R):
print("Run " + str(i + 1) + " of " + str(R))
model = Model(input_size=len(trainX[0]))
model.add_layer(states=2, activation='tanh', fixed_weights=False)
model.add_layer(states=1, activation=None, fixed_weights=1)
# model.display()
# print(trainX.shape)
E, E_test = model.train(trainX,
trainY,
testX,
testY,
ephochs=t_max,
learning_rate=learning_rate)
Es.append(E)
E_tests.append(E_test)
average_E = np.average(Es, axis=0)
from tensorflow.keras.datasets import mnist
from tensorflow.keras.utils import to_categorical
if __name__ == "__main__":
# Load data
(x_train, y_train), (x_test, y_test) = mnist.load_data()
# Flatten input 28 * 28 matrix into a 784 vector
x_train = x_train.reshape(len(x_train), 28 * 28)
x_test = x_test.reshape(len(x_test), 28 * 28)
# Make y into one-hot encoding
y_train = to_categorical(y_train, 10)
y_test = to_categorical(y_test, 10)
# Build model
model = Model()
model.add_layer(in_dim=28 * 28, out_dim=500, activation="tanh")
model.add_layer(out_dim=500, activation="sigmoid")
model.add_layer(out_dim=10, activation="softmax")
# Compile model
model.compile(loss="cross entropy")
# Train model
model.fit(x_train, y_train, batch_size=256, epochs=5)
# Predict and evaluate
loss, accuracy = model.evaluate(x_test, y_test)
print("Total loss for prediction: {}, accuracy: {}".format(loss, accuracy))