How design a autoencoder architecture

Question

I would like to build an autoencoder (CNN) to learn a representation of my data.

I never built such a network and I have some experience in supervised learning (classification).

I would like to know if some good practices in training a classifier is also right for an autoencoder:

1. Does reference architecture exists like ResNet/Inception or something? If not, should I design manually layers?

2. Does transfer learning/fine tuning works for autoencoder (or is it better to train from scratch)?

Answer 1

Yes, there are open source examples. Take a look at here and here. About your second question, yes. There are numerous studies. For instance, take a look at Supervised Representation Learning: Transfer Learning with Deep Autoencoders.