xifengguo / idec Goto Github PK

I'm trying to convert the code into python3 and eras 2.6.0 and getting this error.

ValueError: Layer model_1 expects 1 input(s), but it received 2 input tensors. Inputs received: [<tf.Tensor 'IteratorGetNext:0' shape=(32, 28, 28) dtype=uint8>, <tf.Tensor 'ExpandDims:0' shape=(32, 1) dtype=uint8>]

hello,usps dataset url is invalid ,I can't find it after I serach in google, can you upload it or send it to me ?
Thank you !

Thanks for your work and sharing! There is an error on cluster_acc(Y_pred, Y) function, as
from sklearn.utils.linear_assignment_ import linear_assignment
is depreciated with updated scikit-learn version.

Nice work i have seen the DEC model too and read IDEC paper but i am having some issues in this IDEC code since i want to see how much approved result i get from same dataset of 20_Newsgroup which i have trained on DEC. But using it on IDEC i am getting this error:

    self.n_stacks = self.dims - 1

TypeError: unsupported operand type(s) for -: 'list' and 'int'

And that error makes sense too since 'dims' is a list and we are assigning 'n_stacks' which is an integer by subtracting a list. So it should give type error. But i cannot understand the purpose of this n_stacks as i could not find in paper too please help me as soon as you can.

Thanks,

EDIT: Sorry . It was my mistake i made many changes to original code and that was something from those changes. But now code runs perfectly. Thanks again.

Hello. First of all, I appreciate your wonderful work.

I have one question. How can I implement as unsupervised version?

I think application part is only for the supervised learning with label.

hello, how to visualize the clustering results in your paper?Thank you.

Thanks for your work and sharing! I have carefully read your paper and code, but I have a big doubt.

1. Is there any basis for setting the 'update interval' variable? Why do different datasets correspond to different values? How should I analyze the appropriate values for my own dataset? Can you give me an explanation?
2. About the Autoencoder, what you mean in the paper is that the autoencoder with noise is only used during the pre-training stage, whereas in the IDEC training or testing stage, the autoencoder is only used with undercomplete structure and no noise is added. I want to make sure, do I have an exact understanding?
   Thank you so much!

Hi!

I've tried to download REUTERS data with the get_data.sh script you have provided, but it seems the files are not available there anymore. Do you how I could get that data?

Thank you very much in advance!