tensorflow 学习笔记： recurrent models of visual attention

tf.truncated_normal：

tf.truncated_normal(shape, mean=0.0, stddev=1.0, dtype=tf.float32, seed=None, name=None)tf.truncated_normal(s