python - dropout 中的 keep_prob 值以及通过 dropout 获得最差结果

Question

根据此链接，keep_prob 的值必须在 (0,1] 之间: Tensorflow manual

否则我会得到值错误:

ValueError: If keep_prob is not in (0, 1] or if x is not a floating point tensor.

我将以下代码用于具有一个隐藏层的简单神经网络:

n_nodes_input = len(train_x.columns) # number of input features
n_nodes_hl = 30     # number of units in hidden layer
n_classes = len(np.unique(Y_train_numeric)) 
lr = 0.25
x = tf.placeholder('float', [None, len(train_x.columns)])
y = tf.placeholder('float')
dropout_keep_prob = tf.placeholder(tf.float32)

def neural_network_model(data, dropout_keep_prob):
    # define weights and biases for all each layer
    hidden_layer = {'weights':tf.Variable(tf.truncated_normal([n_nodes_input, n_nodes_hl], stddev=0.3)),
                      'biases':tf.Variable(tf.constant(0.1, shape=[n_nodes_hl]))}
    output_layer = {'weights':tf.Variable(tf.truncated_normal([n_nodes_hl, n_classes], stddev=0.3)),
                    'biases':tf.Variable(tf.constant(0.1, shape=[n_classes]))}
    # feed forward and activations
    l1 = tf.add(tf.matmul(data, hidden_layer['weights']), hidden_layer['biases'])
    l1 = tf.nn.sigmoid(l1)
    l1 = tf.nn.dropout(l1, dropout_keep_prob)
    output = tf.matmul(l1, output_layer['weights']) + output_layer['biases']

    return output

def main():
    prediction = neural_network_model(x, dropout_keep_prob)
    cost = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(labels=y,logits=prediction))
    optimizer = tf.train.AdamOptimizer(lr).minimize(cost)

    sess = tf.InteractiveSession()

    tf.global_variables_initializer().run()
    for epoch in range(1000):
        loss = 0
        _, c = sess.run([optimizer, cost], feed_dict = {x: train_x, y: train_y, dropout_keep_prob: 4.})
        loss += c

        if (epoch % 100 == 0 and epoch != 0):
            print('Epoch', epoch, 'completed out of', 1000, 'Training loss:', loss)
    correct = tf.equal(tf.argmax(prediction,1), tf.argmax(y,1))
    accuracy = tf.reduce_mean(tf.cast(correct, tf.float32), name='op_accuracy')

    writer = tf.summary.FileWriter('graph',sess.graph)
    writer.close()

    print('Train set Accuracy:', sess.run(accuracy, feed_dict = {x: train_x, y: train_y, dropout_keep_prob: 1.}))
    print('Test set Accuracy:', sess.run(accuracy, feed_dict = {x: test_x, y: test_y, dropout_keep_prob: 1.}))
    sess.close()


if __name__ == '__main__':
     main()

如果我在 sess.run 中使用 dropout_keep_prob 范围内的数字 (0,1]，准确度会急剧下降。如果我使用大于 1 的数字，例如 4，准确度会超过 0.9。 一旦我在 tf.nn.dropout() 前面使用shift+tab，这就会被写为描述的一部分:

With probability `keep_prob`, outputs the input element scaled up by
`1 / keep_prob`, otherwise outputs `0`.  The scaling is so that the expected
sum is unchanged.

在我看来，keep_prob 必须大于 1，否则不会丢失任何内容!

底线，我很困惑。我在 dropout 的哪一部分实现了错误，导致我的结果变得最差，keep_drop 的最佳数字是多少？

谢谢

Answer 1

which seems to me that keep_prob has to be greater than 1 otherwise nothing would be dropped!

描述说:

With probability keep_prob, outputs the input element scaled up by 1 / keep_prob, otherwise outputs 0. The scaling is so that the expected sum is unchanged.

这意味着:

• keep_prob用作概率，因此根据定义它应始终位于 [0, 1] 中(超出该范围的数字永远不可能是概率)
• 有概率keep_prob ，输入元素乘以 1 / keep_prob 。因为我们刚刚写了 0 <= keep_prob <= 1 ，师1 / keep_prob总是大于 1.0 (或者恰好是 1.0 如果 keep_prob == 1 )。因此，概率 keep_prob ，一些元素会变得比没有 dropout 时更大
• 有概率1 - keep_prob (描述中的“否则”)，元素设置为 0 。这是 dropout，如果元素设置为 0，则元素将被丢弃。 。如果您设置keep_prob正好1.0 ，这意味着丢弃任何节点的概率变为 0 。所以，如果你想删除一些节点，你应该设置 keep_prob < 1 ，如果您不想删除任何内容，请设置 keep_prob = 1 .

重要说明:您只想在训练期间使用 dropout，而不是在测试期间使用。

If I use a number in range (0,1] for dropout_keep_prob in the sess.run, the accuracy drops drastically.

如果您对测试集执行此操作，或者您的意思是报告训练集的准确性，那么我并不感到惊讶。 Dropout意味着丢失信息，所以确实会丢失准确性。不过，这应该是一种规范化的方式；您在训练阶段故意失去准确性，但希望这会提高泛化能力，从而提高测试阶段(当您不应再使用 dropout 时)的准确性。

If I use a number bigger than 1, like 4, the accuracy goes beyond 0.9.

我很惊讶你竟然能运行这段代码。基于source code ，我不希望它能够运行？