Generative Model Comparison
Chapter 16: Variational Inference and Latent Variables — Generative Model Comparison
From the book
Chapter 16: Variational Inference and Latent Variables. In the chapter mind map this icon labels Generative Model Comparison. The discussion below is excerpted and lightly edited from § Generative Model Comparison in Mathematics for AI and Machine Learning.
While variational inference is a powerful framework for approximate inference, there are other mathematical approaches to generative modeling with latent variables. Two important alternatives are:
- Generative Adversarial Networks (GANs): Use minimax optimization (see Chapter the referenced section, Section the referenced section) rather than variational inference. GANs illustrate how game-theoretic optimization can be applied to generative modeling.
- Normalizing Flows: Use invertible transformations and the change of variables formula (see Chapter the referenced section, Section the referenced section) to create expressive variational families. Normalizing flows can be used within variational inference to create more expressive posterior approximations, or as standalone generative models.
What this drawing shows
What you see. GAN, normalizing-flow, and diffusion columns highlight in turn, contrasting adversarial, invertible, and score-based generation.
In the mind map. Chapter 16 — Generative Model Comparison. See From the book above for definitions, figures, and worked examples.
Where to read next
Read the full definitions, figures, and worked examples in Chapter 16: Variational Inference and Latent Variables — see the mind-map node Generative Model Comparison.