Autoencoders

Unsupervised approach for learning feature vectors from raw data x, without any labels

How can we learn this feature transform from raw data?

Use the features to reconstruct the input data with a decoder

Want features to be lower dimensional than data

Compress input data

After training, throw away decoder and use encode for a downstream task
Pasted image 20241205161914.png

Autoencoders can reconstruct data, and can learn features to initialize a supervised model
Features capture factors of variation in training data
We can’t generate new images from an autoencoder because we don’t know the space of z