# Update output layer weights and biases
dWhy += np.dot(dy, h.T)
dby += dy