I recently learnt about the anamoly detection using autoencoders(specifically denoisinng autoencoders).To train the autoencoders do we need a data having some pattern? or is there some way to train in abnormal data ?Also how we decide that the data is suitable for training autoencoder model.

  • $\begingroup$ Alternatively you could use something like Isolation Forest which allows you to include anomalous data in training. Check out sklearn's implementation and the contamination parameter. $\endgroup$ – Simon Larsson Feb 13 '20 at 11:00
  • $\begingroup$ I think the variational autencoder is superior to the denoising autoencoder, specifically in the ability to generate new data. $\endgroup$ – Victor Ng Mar 14 '20 at 13:56

You need normal data to train on. If you have abnormal instances also, those should be excluded from the training set. Having access to labeled abnormal/normal data is very useful for the validation and testset. Anything that differs from the normal data (as learned by the autoencoder) is considered an anomaly.

If you have a lot of labeled abnormal and normal data, then you can consider using binary classification instead.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.