saved_bias = saved_weights(bias_weights) layer = Dense(2, 2, kernel_initializer=saved_kernel, bias_initializer=saved_bias) sigmoid = Sigmoid() X = np.array([[0, 0], [0, 1], [1, 0], [1, 1]]) y = np.array([[0, 1], [1, 0], [1, 0], [1, 0]]) print("Round 1") z = layer.predict(X) print("Z", z) a = sigmoid.predict(z) print("a", a) error_signal = a - y delta = layer.backprop(error_signal) print("delta", delta) layer.update() print("Round 2") z = layer.predict(X) print("z", z) a = sigmoid.predict(z) print("a", a)