izikgo / anomalydetectiontransformations Goto Github PK

Hi Golan,
Can you tell which of the python file corresponds to your one class classification experiments. I can only see multiclasss experiments and the intuition one on mnist. I was looking for the one for corresponding to one class experiments on cifar10
Thanks

Hi, we download CatVsDog from Kaggle, and find no labels of test set. In your code, the file names of test data are also like cat.0.jpg. Can you share the labels? Or the link to download the labeled dataset. Thank you.

Hey Golan, I was just using your implementations of DSEBM on my own datasets. One problem is that the results (roc_auc) are very unstable. Like in 3 runs, the roc_auc could be 0.31, 0.41, 0.76 where the variance is 0.19! That's too large and in the experiments shown in your paper, the results of DSEBM is quite good and variance the largest 0.069. Do you have any idea of that? Would the reason be my dataset?

@interfaces.legacy_get_updates_support is not defined in wide_residual_network.py

Hi.

I want to run your codes, but this error occurs.

ImportError: cannot import name '_compute_fans' from 'keras.initializers' (/usr/local/lib/python3.7/dist-packages/keras/initializers.py)

I use keras 2.4.3, Python 3.7.10.

How to solve this problem?

@izikgo Can you please provide a requirements txt for a conda environment?