dropout(x, keep_prob, noise_shape=None, seed=None, name=None)

• 函數(shù)作用就是使得矩陣x的一部分（概率大約為keep_prob）變?yōu)?，其余變?yōu)閑lement/keep_prob,
• noise_shape可以使得矩陣x一部分行全為0或者部分列全為0
• 用在tensorflow中使得部分神經(jīng)元隨機(jī)為0不參與訓(xùn)練，如果算法過(guò)擬合了，可以試試這個(gè)辦法。

with tf.Session() as sess:
    d = tf.to_float(tf.reshape(tf.range(1,17),[4,4]))
    sess.run(tf.global_variables_initializer())
    print(sess.run(tf.shape(d)))
    print(sess.run(d[0]))
    
    # 矩陣有一半左右的元素變?yōu)閑lement/0.5,其余為0
    dropout_a44 = tf.nn.dropout(d, 0.5, noise_shape = None)
    result_dropout_a44 = sess.run(dropout_a44)
    print(result_dropout_a44)

    # 行大小相同4，行同為0，或同不為0
    dropout_a41 = tf.nn.dropout(d, 0.5, noise_shape = [4,1])
    result_dropout_a41 = sess.run(dropout_a41)
    print(result_dropout_a41)
    
    # 列大小相同4，列同為0，或同不為0
    dropout_a24 = tf.nn.dropout(d, 0.5, noise_shape = [1,4])
    result_dropout_a24 = sess.run(dropout_a24)
    print(result_dropout_a24)
    #不相等的noise_shape只能為1