tf.nn.sparse_softmax_cross_entropy_with_logits

**def sparse_softmax_cross_entropy_with_logits(_sentinel=None,  # pylint: disable=invalid-name
 labels=None,logits=None,name=None):**
 labels是非稀疏的，例如稀疏矩阵为
 [[ 1.  0.  0.]
 [ 0.  1.  0.]
 [ 0.  0.  1.]]
 shape:[3,3]
转化为非稀疏矩阵为：[0 1 2], shape:[3]
转化方式：yy_=tf.arg_max(y_,1)
A common use case is to have logits of shape `[batch_size, num_classes]` and
  labels of shape `[batch_size]`. But higher dimensions are supported.
logits是未归一化的概率，Unscaled log probabilities of shape
      `[d_0, d_1, ..., d_{r-1}, num_classes]` and dtype `float32` or `float64`.
    name: A name for the operation (optional).
yy_=tf.arg_max(y_,1)
cross_entropy2_step1=tf.nn.sparse_softmax_cross_entropy_with_logits(labels=yy_,logits=logits)
cross_entropy2_step2=tf.reduce_sum(cross_entropy2_step1)


tf.nn.softmax_cross_entropy_with_logits与其主要的区别为labels，这里的
yy_=y_
cross_entropy2_step1=tf.nn.sparse_softmax_cross_entropy_with_logits(labels=yy_,logits=logits)

cross_entropy2_step2=tf.reduce_sum(cross_entropy2_step1)2_step1)