eric-haibin-lin commented on a change in pull request #9625: sparse regression
operators
URL: https://github.com/apache/incubator-mxnet/pull/9625#discussion_r170439324
##########
File path: src/operator/regression_output.cc
##########
@@ -74,12 +79,17 @@ then the squared loss estimated over :math:`n` samples is
defined as
.. note::
Use the LinearRegressionOutput as the final output layer of a net.
+The storage type of ``label`` can be ``default`` or ``csr``
+
+- LinearRegressionOutput(default, default) = default
+- LinearRegressionOutput(default, csr) = default
+
By default, gradients of this loss function are scaled by factor `1/m`, where
m is the number of regression outputs of a training example.
The parameter `grad_scale` can be used to change this scale to `grad_scale/m`.
)code" ADD_FILELINE);
-MXNET_OPERATOR_REGISTER_REGRESSION_BWD(_backward_linear_reg_out,
mshadow_op::minus);
+MXNET_OPERATOR_REGISTER_REGRESSION_BWD(_backward_linear_reg_out,
mshadow_op::minus, true);
Review comment:
Please also update ndarray/sparse.md and symbol/sparse.md
----------------------------------------------------------------
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:
[email protected]
With regards,
Apache Git Services
