[batch-channel-height-width] image = np.transpose(image, (2, 0, ... E1 = np.dot(W2, np.transpose(E2)) ## Eb1 is the error bias for hidden layer.
確定! 回上一頁