boyangumn / dcn-new Goto Github PK

Hello, I am researching about deep clustering and plan to incorporate this work.

However, I am confused about why kmeans is used to replace the updated centers in the code (DCN-New/multi_layer_km.py 1121-1146 lines), which is even not appeared in the old version project (DCN).

Look forward to your reply. Have a good time!

Hi,
I'm interested in the encoded features of your model. I have a training set and a test set. I trained the model using the training set. And I wonder if there is a easy way to encode the test data using the trained model and get the encoded features (just like the model.transform() function in most sklean models) ? The provided code seems to miss this part.
Thanks.

I want to use pendigits to clusters,but i do not how to change the code.I change the data input and the parameters in the paper,but the code run the wrong results. I do not how to change,coule you help me

Hello!
I run your github code ,DCN-NEW,by setting config options "init" as ' ' ,and I pretrain for 250 epoches ,finetune for 50 epoches .But confusingly ,I got the NMI result 0.65 which is not the same as shown in your paper .Will you please give me a guide ?
Best wishes !