Skip to content

Latest commit

 

History

History
26 lines (17 loc) · 1.08 KB

README.md

File metadata and controls

26 lines (17 loc) · 1.08 KB

Deep-Clustering-with-Convolutional-Autoencoders and a simple solution to the clustering problem( img2vec & cosine similarity)

Code from paper : https://xifengguo.github.io/papers/ICONIP17-DCEC.pdf

The implementation of this network was also taken as a basis : https://github.com/michaal94/torch_DCEC

About CNN

This network uses the (Encoder x Decoder) with params [32,64,128] architecture, but there is a clustering layer between the layers.

Clustering Loss

Train

Encoder_Decoder :

  • Input : Tensor(batch_size, channels , (size,size)). Default params : [16, 3, 128, 128]
  • Loss Function : torch.nn.MSE() output = model(x_train) loss = Loss Function(output, x_train)

Other

This is just the first simple implementation of DCEC, improved implementations can be viewed from the links above.

  • $ docker build -t $IMAGE_TAG -f ./Dockerfile ./