classificationLayer cannot output one-hot encoded matrix?

5 次查看（过去 30 天）
Xiaolong
Xiaolong 2019-7-10
评论： Sanjana Sankar 2019-7-30
openExample('nnet/ConstructAndTrainAConvolutionalNeuralNetworkExample')
In this demo, the output of the net was a categorical vector.
But in most of the cnn tutorials, the output should be one-hot encoded. Why does matlab use a categorical vector instead?
How to feed a one-hot encoded matrix to the net? The function trainNetwork() does not accept this kind of Y data.
Is the classificationLayer itself implemented one-hot encoding?
  2 个评论
Maria Duarte Rosa
Maria Duarte Rosa 2019-7-11
Hi Xiaolong, Thanks for your question. Is there a particular reason (apart from seeing in other cnn tutorials) why you think one-hot encoded outputs are better over categorical vectors? Do you have a particular use case? I'm just trying to understand your feedback/question.
Sanjana Sankar
Sanjana Sankar 2019-7-30
Hi can someone tell me how to do one hot encoding in MATLAB to feed to a network?

请先登录，再进行评论。

请先登录，再回答此问题。

回答（1 个）

Himanshu tripathi
Himanshu tripathi 2019-7-10
Hey Xiaolong,
Why do you want to feed one hot encoded Y to the net?
The function trainNetwork() does not accept one hot encoded matrix and it's expected behaviour as per documentation.

类别

在 Help Center 和 File Exchange 中查找有关 Image Data Workflows 的更多信息

标签

产品


版本

R2018a

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by