October 2018
Beginner
362 pages
9h 32m
English
Like standard autoencoders, VAEs utilize the same encoder/decoder framework, but, that aside, they are mathematically different from their namesake. VAEs take a probabilistic perspective in terms of guiding the network:

Both our encoder and decoder networks are generating distributions from their input data. The encoder generates a distribution from its training data, Z, which then becomes the input distribution for the decoder. The decoder takes this distribution, Z, and tries to replicate the original distribution, X, from it.
Read now
Unlock full access