singa-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "ASF subversion and git services (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (SINGA-184) Add Cross Entropy loss computation
Date Mon, 13 Jun 2016 09:20:21 GMT

    [ https://issues.apache.org/jira/browse/SINGA-184?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15327043#comment-15327043
] 

ASF subversion and git services commented on SINGA-184:
-------------------------------------------------------

Commit ec17acab49d595fdc48b2dae6f71901b5a4c8191 in incubator-singa's branch refs/heads/dev
from [~flytosky]
[ https://git-wip-us.apache.org/repos/asf?p=incubator-singa.git;h=ec17aca ]

SINGA-184 Add Cross Entropy loss computation

Update softmaxcrossentropy layer to support both cpp and cuda devices;

Fix bugs from crossentropy fwd and bwd; need the cuda version exp();


> Add Cross Entropy loss computation
> ----------------------------------
>
>                 Key: SINGA-184
>                 URL: https://issues.apache.org/jira/browse/SINGA-184
>             Project: Singa
>          Issue Type: New Feature
>            Reporter: Zheng Kaiping
>
> Implement Cross Entropy loss.
> Cross Entropy loss accepts the prediction from the neural net and the ground truth from
the training data. It outputs a scalar loss value for each data sample and computes the gradient
of the loss value w.r.t the prediction value through back propagation.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message