TensorFlow get_variable argmax equal cast reduce_mean函数示例

代码如下：

import tensorflow as tf

var_a_op = tf.get_variable(name='a',
                    shape=[3, 4],
                    dtype=tf.float32,
                    initializer=tf.random_uniform_initializer(minval=-1, maxval=1))

var_b_op = tf.get_variable(name='b',
                    shape=[3, 4],
                    dtype=tf.float32,
                    initializer=tf.random_uniform_initializer(minval=-1, maxval=1))

max_a_op = tf.argmax(var_a_op, 1)
max_b_op = tf.argmax(var_b_op, 1)

equal_op = tf.equal(max_a_op, max_b_op)
cast_op = tf.cast(equal_op, "float")
mean_op = tf.reduce_mean(cast_op)

init_op = tf.initialize_all_variables()

with tf.Session() as sess:
    sess.run(init_op)
    print(sess.run(var_a_op))
    print(sess.run(var_b_op))
    print(sess.run(max_a_op))
    print(sess.run(max_b_op))
    print(sess.run(equal_op))
    print(sess.run(cast_op))
    print(sess.run(mean_op))

输出如下：

[[ 0.86882663  0.83623457  0.18949676  0.1206305 ]
 [ 0.26341105 -0.96493745  0.42790508  0.79426742]
 [ 0.07995605 -0.54766798  0.81907773 -0.13814092]]
[[ 0.9508462   0.69852757 -0.42363882 -0.83991981]
 [ 0.18524432  0.28357625 -0.33817434 -0.10542083]
 [ 0.32140636  0.00276613  0.57352829 -0.22276378]]
[0 3 2]
[0 1 2]
[ True False  True]
[ 1.  0.  1.]
0.666667