[GitHub] szha closed pull request #9404: Use a better formula to calculate sigmoid_bce and logistic_loss

[GitBox <gi...@apache.org>]

szha closed pull request #9404: Use a better formula to calculate sigmoid_bce and logistic_loss
URL: https://github.com/apache/incubator-mxnet/pull/9404
 
 
   

This is a PR merged from a forked repository.
As GitHub hides the original diff on merge, it is displayed below for
the sake of provenance:

As this is a foreign pull request (from a fork), the diff is supplied
below (as it won't show otherwise due to GitHub magic):

diff --git a/python/mxnet/gluon/loss.py b/python/mxnet/gluon/loss.py
index 435230ef53..2be43981a6 100644
--- a/python/mxnet/gluon/loss.py
+++ b/python/mxnet/gluon/loss.py
@@ -229,8 +229,8 @@ def __init__(self, from_sigmoid=False, weight=None, batch_axis=0, **kwargs):
     def hybrid_forward(self, F, pred, label, sample_weight=None):
         label = _reshape_like(F, label, pred)
         if not self._from_sigmoid:
-            max_val = F.relu(-pred)
-            loss = pred - pred*label + max_val + F.log(F.exp(-max_val)+F.exp(-pred-max_val))
+            # We use the stable formula: max(x, 0) - x * z + log(1 + exp(-abs(x)))
+            loss = F.relu(pred) - pred * label + F.Activation(-F.abs(pred), act_type='softrelu')
         else:
             loss = -(F.log(pred+1e-12)*label + F.log(1.-pred+1e-12)*(1.-label))
         loss = _apply_weighting(F, loss, self._weight, sample_weight)
@@ -635,8 +635,8 @@ class LogisticLoss(Loss):
 
     Inputs:
         - **pred**: prediction tensor with arbitrary shape.
-        - **label**: truth tensor with values -1 or 1. Must have the same size
-          as pred.
+        - **label**: truth tensor with values -1/1 (label_format is 'signed')
+          or 0/1 (label_format is 'binary'). Must have the same size as pred.
         - **sample_weight**: element-wise weighting tensor. Must be broadcastable
           to the same shape as pred. For example, if pred has shape (64, 10)
           and you want to weigh each sample in the batch separately,
@@ -655,9 +655,10 @@ def __init__(self, weight=None, batch_axis=0, label_format='signed', **kwargs):
 
     def hybrid_forward(self, F, pred, label, sample_weight=None):
         label = _reshape_like(F, label, pred)
-        if self._label_format == 'binary':
-            label = 2 * label - 1  # Transform label to be either -1 or 1
-        loss = F.log(1.0 + F.exp(-pred * label))
+        if self._label_format == 'signed':
+            label = (label + 1.0) / 2.0  # Transform label to be either 0 or 1
+        # Use a stable formula in computation
+        loss = F.relu(pred) - pred * label + F.Activation(-F.abs(pred), act_type='softrelu')
         loss = _apply_weighting(F, loss, self._weight, sample_weight)
         return F.mean(loss, axis=self._batch_axis, exclude=True)
 
diff --git a/tests/python/unittest/test_loss.py b/tests/python/unittest/test_loss.py
index e044df0705..9fb303352b 100644
--- a/tests/python/unittest/test_loss.py
+++ b/tests/python/unittest/test_loss.py
@@ -97,6 +97,14 @@ def test_bce_loss():
             eval_metric=mx.metric.Loss(), optimizer='adam',
             initializer=mx.init.Xavier(magnitude=2))
     assert mod.score(data_iter, eval_metric=mx.metric.Loss())[0][1] < 0.01
+    # Test against npy
+    data = mx.random.uniform(-5, 5, shape=(10,))
+    label = mx.random.uniform(0, 1, shape=(10,))
+    mx_bce_loss = Loss(data, label).asnumpy()
+    prob_npy = 1.0 / (1.0 + np.exp(-data.asnumpy()))
+    label_npy = label.asnumpy()
+    npy_bce_loss = - label_npy * np.log(prob_npy) - (1 - label_npy) * np.log(1 - prob_npy)
+    assert_almost_equal(mx_bce_loss, npy_bce_loss)
 
 def test_bce_equal_ce2():
     N = 100
@@ -107,6 +115,15 @@ def test_bce_equal_ce2():
     label = mx.nd.round(mx.random.uniform(0, 1, shape=(N, 1)))
     assert_almost_equal(loss1(out1, label).asnumpy(), loss2(out2, label).asnumpy())
 
+def test_logistic_loss_equal_bce():
+    N = 100
+    loss_binary = gluon.loss.LogisticLoss(label_format='binary')
+    loss_signed = gluon.loss.LogisticLoss(label_format='signed')
+    loss_bce = gluon.loss.SigmoidBCELoss(from_sigmoid=False)
+    data = mx.random.uniform(-10, 10, shape=(N, 1))
+    label = mx.nd.round(mx.random.uniform(0, 1, shape=(N, 1)))
+    assert_almost_equal(loss_binary(data, label).asnumpy(), loss_bce(data, label).asnumpy())
+    assert_almost_equal(loss_signed(data, 2 * label - 1).asnumpy(), loss_bce(data, label).asnumpy())
 
 def test_kl_loss():
     np.random.seed(1234)


 

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
users@infra.apache.org


With regards,
Apache Git Services