liveBook · Manning

Likelihood-based generative model

As explained in Machine Learning Algorithms in Depth

A likelihood-based generative model represents a data distribution with a likelihood function and learns by fitting that likelihood.

A likelihood-based generative model represents the data distribution with a likelihood function and learns by fitting that likelihood. It contrasts with generative adversarial networks, which use adversarially trained generator and discriminator networks to approximate the data distribution. section 11.5

Variational autoencoders form the most popular subclass. An encoder maps data to a latent representation, and a decoder maps that representation back to a reconstruction. Beta-VAEs modify this structure to emphasize disentangled latent factors. section 11.5

Related likelihood-based approaches include:

  • Variational autoencoders use latent representations and reconstruction. section 11.5
  • Beta-VAEs emphasize discovering disentangled latent factors. section 11.5
  • Diffusion models use variational inference and reverse a noise-adding process to generate samples matching the target distribution. section 11.5

How Machine Learning Algorithms in Depth covers it

How deep it goes
Defines likelihood modeling and identifies VAE and diffusion structures, but stops short of deriving objectives or implementation steps.
Examples
Uses VAEs, beta-VAEs, and diffusion models as representative likelihood-based examples.
What you can do after
Evaluate distinctions among likelihood-based, adversarial, and diffusion generative-model architectures.
How it connects
Connects likelihood functions with variational autoencoders, latent representations, variational inference, diffusion models, and generative adversarial networks.
Cover of Machine Learning Algorithms in Depth
liveBookmanning.com
Read Machine Learning Algorithms in Depth on liveBook Get the book at manning.com

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Machine Learning Algorithms in Depth. The section links point to the parts of the book it draws on.