machine-learning - Tensorflow - 使用 tf.losses.hinge_loss 导致形状不兼容错误

Question

我当前使用sparse_softmax_cross_entropy的代码工作正常。

loss_normal = (
    tf.reduce_mean(tf.losses
                   .sparse_softmax_cross_entropy(labels=labels,
                                                 logits=logits,
                                                 weights=class_weights))
    )

但是，当我尝试使用hinge_loss时:

loss_normal = (
    tf.reduce_mean(tf.losses
                   .hinge_loss(labels=labels,
                               logits=logits,
                               weights=class_weights))
    )

它报告了一个错误:

ValueError: Shapes (1024, 2) and (1024,) are incompatible

该错误似乎源自 losses_impl.py 文件中的此函数:

  with ops.name_scope(scope, "hinge_loss", (logits, labels)) as scope:
    ...
    logits.get_shape().assert_is_compatible_with(labels.get_shape())
    ...

我修改了代码如下，只提取 logits 张量的 1 列:

loss_normal = (
    tf.reduce_mean(tf.losses
                   .hinge_loss(labels=labels,
                               logits=logits[:,1:],
                               weights=class_weights
                               ))
    )

但是还是报类似的错误:

ValueError: Shapes (1024, 1) and (1024,) are incompatible.

有人可以帮忙指出为什么我的代码可以在 sparse_softmax_cross_entropy 损失下正常工作，但在 hinge_loss 上却不能正常工作吗？

Answer 1

张量labels的形状为[1024]，张量logits的形状为[1024, 2]形状。这适用于 tf.nn.sparse_softmax_cross_entropy_with_logits :

• labels: Tensor of shape [d_0, d_1, ..., d_{r-1}] (where r is rank of labels and result) and dtype int32 or int64. Each entry in labels must be an index in [0, num_classes). Other values will raise an exception when this op is run on CPU, and return NaN for corresponding loss and gradient rows on GPU.
• logits: Unscaled log probabilities of shape [d_0, d_1, ..., d_{r-1}, num_classes] and dtype float32 or float64.

但是tf.hinge_loss要求不同:

• labels: The ground truth output tensor. Its shape should match the shape of logits. The values of the tensor are expected to be 0.0 or 1.0.
• logits: The logits, a float tensor.

您可以通过两种方式解决此问题:

• 将标签 reshape 为 [1024, 1] 并仅使用一行 logits，就像您所做的那样 - logits[:,1: ]:

  labels = tf.reshape(labels, [-1, 1])
  hinge_loss = (
      tf.reduce_mean(tf.losses.hinge_loss(labels=labels,
                                          logits=logits[:,1:],
                                          weights=class_weights))
      )
  

  我认为您还需要以同样的方式 reshape class_weights。

• 通过tf.reduce_sum使用所有学习的logits特征，这将生成一个平坦的(1024,)张量:

  logits = tf.reduce_sum(logits, axis=1)
  hinge_loss = (
      tf.reduce_mean(tf.losses.hinge_loss(labels=labels,
                                          logits=logits,
                                          weights=class_weights))
      )
  

  这样您就不需要 reshape 标签或class_weights。