How to design a Deep Recurrent Denoising Autoencoder

[Quazi Marufur Rahman]

Hi,

Based on this and this paper, I want to design a Deep Recurrent Denoising Autoencoder. I can design an autoencoder based on the examples provided by Caffe. However,

I don't know how to implement the recurrent attribute (The output of middle layer should be used alongside with the next input). Is there any prototxt available as example? Please check the attachment to know about what I am intending to do.

Would you please suggest me, how can I do this?

Thanks in advance.

[Jan]

I don't think vanilla caffe supports recurrency (at least I am not aware of it). But there are recurrent forks. Try one of these, maybe?

Jan

[Quazi Marufur Rahman]

Could you please point me one? I haven't found anything.

[Jan]

http://lmgtfy.com/?q=caffe+recurrent