Testing and training function
def CNN_training(image, label, layers, alpha=2.0):
# Forward step
output, loss, accuracy = CNN_forward(image, label, layers)
# Initial gradient
gradient = np.zeros(10)
gradient[label] = -1/output[label]
# Backprop step
gradient_back = CNN_backprop(gradient, layers, alpha)
return loss, accuracy
def CNN_testing(image, label, layers):
# Perform forward propagation
for layer in layers:
image = layer.forward(image)
# Compute the loss and accuracy
loss = -np.log(image[label])
accuracy = 1 if np.argmax(image) == label else 0
return loss, accuracy
Main function
def main():
# Load training data
(X_train, y_train), (X_test, y_test) = tf.keras.datasets.mnist.load_data()
X_train_full = X_train[:5000]
y_train_full = y_train[:5000]
# Define the network
layers = [
ConvolutionLayer(16, 3), # layer with 16 3x3 filters, output (26, 26, 16)
MaxPoolingLayer(2), # pooling layer 2x2, output (13, 13, 16)
SoftmaxLayer(13 * 13 * 16, 10) # softmax layer with 5*5*32 input and 10 output
]
learning_rate = 0.05
batch_size = 10
epochs = 4
accuracy_results = []
time_results = []
average_loss_results = []
Loop for getting the result for 4 epoch
for run in range(1):
print(f"Run {run + 1}")
accuracy_list = []
average_loss_list = []
start_time = time.time()
for epoch in range(epochs):
print('Epoch {} ->'.format(epoch + 1))
# Shuffle training data
permutation = np.random.permutation(len(X_train_full))
X_train = X_train_full[permutation]
y_train = y_train_full[permutation]
# Training the CNN
loss = 0
accuracy = 0
for i, (image, label) in enumerate(zip(X_train, y_train)):
if i % batch_size == 0:
print("Step {}. For the last {} steps: average loss {}, accuracy {}".format(
i + 1, batch_size, loss / batch_size, accuracy / batch_size))
loss = 0
accuracy = 0
batch_loss, batch_accuracy = CNN_training(image, label, layers, learning_rate)
loss += batch_loss
accuracy += batch_accuracy
# Append accuracy and average loss to lists
accuracy_list.append(batch_accuracy)
average_loss_list.append(batch_loss / batch_size)
Calculate for the computational time in the loop
end_time = time.time()
total_time = end_time - start_time
print(f"Total time taken for training: {total_time:.2f} seconds")
average_accuracy = sum(accuracy_list) / len(accuracy_list)
print("Average Accuracy: {:.2%}".format(average_accuracy))
# Save the results from each run
accuracy_results.append(average_accuracy)
time_results.append(total_time)
average_loss_results.append(average_loss_list)
Plotting the graph
# Plot accuracy, time, and average loss
plt.figure(figsize=(12, 4))
# Average loss plot
plt.subplot(1, 3, 3)
for avg_loss in average_loss_results:
plt.plot(x_values, avg_loss) # Use x_values as the x-axis values
plt.xlabel('Epoch')
plt.ylabel('Average Loss')
plt.title('Average Loss during Training')
plt.tight_layout()
plt.show()
What is the wrong on the graph as the output is not show any plotting on it? I want to get two graph which are training average loss per epoch and testing average loss per epoch
