What are the advantages/disadvantages of using Autoencoders over CNNs for image search?

1

1

I've seen both of these techniques be used for image search. One difference I can think of is that autoencoders don't rely on labeled data. I'm not sure, but it seems logical therefore that they can possibly generate more discriminatory dimensions for the final vector-representation, given that you're no longer bound by the classifications from the labels.

For my particular problem, I have labeled data, which is why I'm stuck between the 2. This is for non-CNN autoencoders.

What are the advantages/disadvantages of using Autoencoders over CNNs for image search?

aces

Posted 2016-06-17T01:18:57.753

Reputation: 111

1

Auto-encoders and CNNs are not mutually exclusive architectures. You can have an auto-encoding CNN. E.g. https://swarbrickjones.wordpress.com/2015/04/29/convolutional-autoencoders-in-pythontheanolasagne/ - so for clarification is your question about difference between using a classifier vs auto-encoder for search? Is it important to you to compare a CNN with a non-CNN autoencoder (probably no-one would use a non-CNN autoencoder for image search nowadays)?

– Neil Slater – 2016-06-17T05:55:06.170

Yes, non-CNN autoencoders. What you mention problem might result in me posting a new question though, because I read negative things about CNN autoencoders, but I'm rather new to ML in general, so I wouldn't know. I found autoencoders being suggested for image search here: http://deeplearning4j.org/neuralnetworktable

– aces – 2016-06-17T22:26:52.140

No answers