!6832 Add BCELoss

Merge pull request !6832 from wanyiming/add_BCELoss

mindspore/nn/loss/__init__.py

@@ -20,8 +20,8 @@ It shows how well the model works on a dataset and the optimization target which
 """
 
 from .loss import L1Loss, MSELoss, SmoothL1Loss, \
-    SoftmaxCrossEntropyWithLogits, CosineEmbeddingLoss
+    SoftmaxCrossEntropyWithLogits, BCELoss, CosineEmbeddingLoss
 
 __all__ = ['L1Loss', 'MSELoss', 'SmoothL1Loss',
-           'SoftmaxCrossEntropyWithLogits',
+           'SoftmaxCrossEntropyWithLogits', 'BCELoss',
            'CosineEmbeddingLoss']

mindspore/nn/loss/loss.py

@@ -262,6 +262,67 @@ class SoftmaxCrossEntropyWithLogits(_Loss):
         return self.get_loss(x)
 
 
+class BCELoss(_Loss):
+    r"""
+    BCELoss creates a criterion to measure the Binary Cross Entropy between the true labels and predicted labels.
+
+    Note:
+        Set the predicted labels as :math:`x`, true labels as :math:`y`, the output loss as :math:`\ell(x, y)`.
+        Let,
+
+        .. math::
+            L = \{l_1,\dots,l_N\}^\top, \quad
+            l_n = - w_n \left[ y_n \cdot \log x_n + (1 - y_n) \cdot \log (1 - x_n) \right]
+
+        Then,
+
+        .. math::
+            \ell(x, y) = \begin{cases}
+            L, & \text{if reduction} = \text{`none';}\\
+            \operatorname{mean}(L), & \text{if reduction} = \text{`mean';}\\
+            \operatorname{sum}(L),  & \text{if reduction} = \text{`sum'.}
+            \end{cases}
+
+    Args:
+        weight (Tensor, optional): A rescaling weight applied to the loss of each batch element.
+            And it must have same shape and data type as `inputs`. Default: None
+        reduction (str): Specifies the reduction to be applied to the output.
+            Its value must be one of 'none', 'mean', 'sum'. Default: 'none'.
+
+    Inputs:
+        - **inputs** (Tensor) - The input Tensor. The data type must be float16 or float32.
+        - **labels** (Tensor) - The label Tensor which has same shape and data type as `inputs`.
+
+    Outputs:
+        Tensor or Scalar, if `reduction` is 'none', then output is a tensor and has the same shape as `inputs`.
+        Otherwise, the output is a scalar. default: 'none'
+
+    Examples:
+        >>> weight = Tensor(np.array([[1.0, 2.0, 3.0], [4.0, 3.3, 2.2]]), mindspore.float32)
+        >>> loss = nn.BCELoss(weight=weight, reduction='mean')
+        >>> inputs = Tensor(np.array([[0.1, 0.2, 0.3], [0.5, 0.7, 0.9]]), mindspore.float32)
+        >>> labels = Tensor(np.array([[0, 1, 0], [0, 0, 1]]), mindspore.float32)
+        >>> loss(inputs, labels)
+    """
+
+    def __init__(self, weight=None, reduction='none'):
+        super(BCELoss, self).__init__()
+        self.binary_cross_entropy = P.BinaryCrossEntropy(reduction=reduction)
+        self.weight_one = weight is None
+        if not self.weight_one:
+            self.weight = weight
+        else:
+            self.ones = P.OnesLike()
+
+    def construct(self, inputs, labels):
+        if self.weight_one:
+            weight = self.ones(inputs)
+        else:
+            weight = self.weight
+        loss = self.binary_cross_entropy(inputs, labels, weight)
+        return loss
+
+
 @constexpr
 def _check_reduced_shape_valid(ori_shape, reduced_shape, axis, cls_name):
     validator.check_reduce_shape(ori_shape, reduced_shape, axis, cls_name)

tests/ut/python/nn/test_loss.py

@@ -53,6 +53,34 @@ def test_SoftmaxCrossEntropyWithLogits_reduce():
     loss(logits, labels)
 
 
+def test_BCELoss():
+    """ test_BCELoss """
+    loss = nn.BCELoss()
+
+    inputs_data = Tensor(np.array([[0.1, 0.2, 0.3], [0.5, 0.7, 0.9]]).astype(np.float32))
+    target_data = Tensor(np.array([[0, 1, 0], [0, 0, 1]]).astype(np.float32))
+    loss(inputs_data, target_data)
+
+
+def test_BCELoss_reduce():
+    """ test_BCELoss """
+    loss = nn.BCELoss(reduction='mean')
+
+    inputs_data = Tensor(np.array([[0.1, 0.2, 0.3], [0.5, 0.7, 0.9]]).astype(np.float32))
+    target_data = Tensor(np.array([[0, 1, 0], [0, 0, 1]]).astype(np.float32))
+    loss(inputs_data, target_data)
+
+
+def test_BCELoss_weight():
+    """ test_BCELoss """
+    weight = Tensor(np.array([[1.0, 2.0, 3.0], [2.2, 2.6, 3.9]]).astype(np.float32))
+    loss = nn.BCELoss(weight=weight)
+
+    inputs_data = Tensor(np.array([[0.1, 0.2, 0.3], [0.5, 0.7, 0.9]]).astype(np.float32))
+    target_data = Tensor(np.array([[0, 1, 0], [0, 0, 1]]).astype(np.float32))
+    loss(inputs_data, target_data)
+
+
 def test_cosine_embedding_loss():
     """ test CosineEmbeddingLoss """
     loss = nn.CosineEmbeddingLoss()