我想知道為什么我的神經(jīng)網(wǎng)絡(luò)不起作用。我想說(shuō)我對(duì)此提出了類似的問(wèn)題，但我仍然有一些我不明白的事情......代碼：import numpy as npinputs = np.array([    [[0],[0]],    [[1],[0]],    [[0],[1]],    [[1],[1]]])expected_output = np.array([    [0],    [1],    [1],    [0]])epochs = 100lr = 0.2hidden_weights = np.array([    [0.2, 0.3],    [0.4, 0.5]])hidden_bias = np.array([[0.3], [0.6]])output_weights = np.array([[0.6, 0.7]])output_bias = np.array([[0.5]])def sigmoid(z):    return 1/(1+np.exp(-z))def sigmoid_derivative(z):    return z * (1.0-z)for _ in range(epochs):    for index, input in enumerate(inputs):        hidden_layer_activation = np.dot(hidden_weights, input)        hidden_layer_activation += hidden_bias        hidden_layer_output = sigmoid(hidden_layer_activation)        output_layer_activation = np.dot(output_weights, hidden_layer_output)        output_layer_activation += output_bias        predicted_output = sigmoid(output_layer_activation)        #Backpropagation        output_errors = expected_output[index] - predicted_output        hidden_errors = output_weights.T.dot(output_errors)        d_predicted_output = output_errors * sigmoid_derivative(predicted_output)        d_hidden_layer = hidden_errors * sigmoid_derivative(hidden_layer_output)        output_weights += np.dot(d_predicted_output, hidden_layer_output.T) * lr        hidden_weights += np.dot(d_hidden_layer, input.T) * lr        output_bias += np.sum(d_predicted_output) * lr        hidden_bias += np.sum(d_hidden_layer) * lr# NOW THE TESTING,I pass 2 input neurons. One with value 1 and value 1test = np.array([    [[1], [1]]])我已經(jīng)測(cè)試了前饋傳播，它工作正常。錯(cuò)誤似乎很好。我認(rèn)為更新權(quán)重是問(wèn)題，但更新權(quán)重有正確的公式。這段代碼來(lái)自《制作你自己的神經(jīng)網(wǎng)絡(luò)》一書(shū)，它與我使用的幾乎相同：self.who += self.lr * numpy.dot((output_errors * final_outputs * (1.0 - final_outputs)), numpy.transpose(hidden_outputs))目前我當(dāng)時(shí)只轉(zhuǎn)發(fā) 2 個(gè)神經(jīng)元的 1 個(gè)輸入并計(jì)算錯(cuò)誤。我非常希望它保持這種狀態(tài)，而不是一遍又一遍地轉(zhuǎn)發(fā)整個(gè)測(cè)試數(shù)據(jù)。有什么辦法可以做到嗎？先感謝您 ：）