Mailing-List: contact commits-help@mxnet.incubator.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@mxnet.incubator.apache.org
From: GitBox <git@apache.org>
To: commits@mxnet.apache.org
Subject: [GitHub] hwjwayne opened a new issue #8933: How to train data with multi-class label
Message-ID: <151238290960.24942.15039133488869586.gitbox@gitbox.apache.org>
archived-at: Mon, 04 Dec 2017 10:21:51 -0000

hwjwayne opened a new issue #8933: How to train data with multi-class label
URL: https://github.com/apache/incubator-mxnet/issues/8933
 
 
   There are 600 classes in the image dataset, each image belongs to one or more classes. How can I train such dataset?
   
   **Is it a correct choice:**
   Present image label by 600*1 binary vector, which contains one or more "1".
   Bulid network with 600 outputs, and activate them with LogisticRegressionOutput()  (BTW, Will mxnet set the loss function to logistic loss automatically?)
   
   I obtain a 0.49% mAP on this dataset, by it's reported 36.1% in the paper. (Parameters are set the same as the paper)

----------------------------------------------------------------
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