search

lukasruff / Deep-SVDD-PyTorch

A PyTorch implementation of the Deep SVDD anomaly detection method
MIT License
698 stars 197 forks source link

about results(Average AUCs in % with StdDevs (over 10 seeds) ) #17

Open tommying opened 4 years ago

tommying commented 4 years ago

Hi, I am mystified by this (Average AUCs in % with StdDevs (over 10 seeds) ).

Do you mean to get the average results over 10 seeds. If then, what should be the range of seeds? (1 - 10 ? or 10 -20 ?). The results varied widely, because of different seed settings. Can you help me with this question? Thank you.

dimimal commented 4 years ago

I have the same question as well! If someone from the team could provide any help, that would be great!

omid-ghozatlou commented 2 years ago

This is also my question and I think its related to collect_results.py in the utils folder. and I have another question: If we set seed to any number except -1, the result must not change by rerunning. however when I use preparing the results of both pretraining and training change widely. it worth mentioning that if you set pretraining False, the network is trained without any pre-trained weights and the result of many times training are the same for the specific seed value. (of course except -1) Anybody knows why seed does not work for autoencoder?

omid-ghozatlou commented 2 years ago

According to collect_results.py range of seed is 1-10

Longchao1047027852 commented 1 year ago

I also have such problems. Should the author publish the seeds, otherwise the training results may deviate due to different seeds.