[
https://issues.apache.org/jira/browse/SINGA-91?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14962417#comment-14962417
]
ASF subversion and git services commented on SINGA-91:
------------------------------------------------------
Commit ef9a11136538713e1f707ebb87a0ae07bde764d2 in incubator-singa's branch
refs/heads/master from [~flytosky]
[ https://git-wip-us.apache.org/repos/asf?p=incubator-singa.git;h=ef9a111 ]
SINGA-91 - Add SoftmaxLayer and ArgSortLayer
SoftmaxLayer applies the Softmax function against its source layer to compute
its probability distribution over all labels.
ArgSortLayer sorts labels based on their scores (e.g., probability) in
descending order.
Configuration for ArgSortLayer is like argsort_conf{ topk: 1}. Topk results
will be extracted.
Connecting ArgSortLayer to a CSVOutputLayer, we can dump the topk labels
of each instance into one line.
> Add SoftmaxLayer and ArgSortLayer
> ---------------------------------
>
> Key: SINGA-91
> URL: https://issues.apache.org/jira/browse/SINGA-91
> Project: Singa
> Issue Type: New Feature
> Reporter: wangwei
>
> These two layers are added to get prediction results (e.g., predicted
> labels), which can be output by CSVOutputLayer.
> SoftmaxLayer applies the Softmax function against its source layer to compute
> its probability distribution over all labels.
> ArgSortLayer sorts labels based on their scores (e.g., probability) in
> descending order. It has a configuration field *argsort_conf*,
> {code}
> message ArgSortProto {
> optional int32 topk = 1 [default = 1];
> }
> {code}
> Topk labels will be kept.
> ArgSortLayer cannot be used in the training phase, because it does not
> implement the ComputeGradient function. It is used merely for extracting
> prediction results. If it connects to a CSVOutputLayer, then topk labels will
> be dumped into a csv file, topk labels per line.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
