Learn Before
Concept

Markov Chain for Generalized Denoising Autoencoders

Each step consists in (a) injecting noise via corruption process C in state x, yielding ˜x, (b) encoding it with function f, yielding h=f(˜x), (c) decoding the result with function g, yielding parameters ω for the reconstruction distribution, and (d) given ω, sampling a new state from the reconstruction distribution p( x | ω=g(f(˜x)) ).

Image 0

0

1

Updated 2021-07-29

Tags

Data Science