mesuvash / nnrec Goto Github PK

Hi,
I have worked on applying deep learning into collaborative filtering, and read your paper very carefully.

However, I cannot reproduce the result on MovieLens 10M data, which I think is lack of optimal parameter setting. Currently, I am using the default parameter in code which might be best for MovieLens 1M, and I got RMSE 0.8937 on 10M data.

Could you please share the one on 10M data?

Could you please point to the test scripts for the model?

Hi Suvash,

I have read your AutoRec paper, it's very interesting paper.

I have implemented Deep AutoRec with Tensorflow.

I was wondering which hyper parameter setting for training deeper AutoRec (greedy train and end-to-end train): optimizer, regularization, batch_size...

Thanks
Hung

Hi,

While running the autorec code for my own custom data I'm getting the following error: ValueError: too many values to unpack. What do you think can be done?

The complete error:

Traceback (most recent call last):                         ]   0% ETA:  --:--:--
  File "learner.py", line 58, in <module>
    train(config_path)
  File "learner.py", line 20, in train
    shape=shape)
  File "/home/soumendra/NNRec/dataUtils/data.py", line 103, in loadTrainTest
    d.import_ratings(train_path, shape)
  File "/home/soumendra/NNRec/dataUtils/data.py", line 37, in import_ratings
    userid, itemid, rating = line.split()
ValueError: too many values to unpack

I can send you the data if you'd like.

Hi,

I have been reading the AutoRec paper, which I found very interesting! I was wondering: what are the very best parameter settings that allowed to reach the reported good results?

I was wondering which learning rate, which normalization coefficient and which bottleneck size has been used..

thanks in advance! :)
-Francesco