Opening Variational Inference, Part 1 (CS 285, Lecture 18)
Variational Inference, Part 1 (CS 285, Lecture 18)
- Length
- 20 min
- Level
- Graduate
- Recorded
- UC Berkeley2020
Chapter 1 of 3: Before the lecture
Probabilistic latent variable models represent a complex distribution over data as a product of simpler ones, with unobserved variables integrated out. Levine opens the first part of the CS 285 variational inference unit by separating conditional from unconditional probabilistic models. He then writes and shows that fitting it by maximum likelihood requires the log of an intractable integral. The fix he builds is the expected log likelihood, an expectation over the posterior that can be estimated with samples. Computing that posterior remains the open problem, which motivates variational methods and amortized inference. The material assumes multivariate normal distributions, basic neural network training and the earlier course lectures.
Your comment is kept while you sign up.Sign in
Shows why maximum likelihood stalls on latent variable models and swaps in an expected log likelihood.
Probabilistic latent variable models represent a complex distribution over data as a product of simpler ones, with unobserved variables integrated out. Levine opens the first part of the CS 285 variational inference unit by separating conditional from unconditional probabilistic models. He then writes p(x) = ∫ p(x | z) p(z) dz and shows that fitting it by maximum likelihood requires the log of an intractable integral. The fix he builds is the expected log likelihood, an expectation over the posterior that can be estimated with samples. Computing that posterior remains the open problem, which motivates variational methods and amortized inference. The material assumes multivariate normal distributions, basic neural network training and the earlier course lectures.
Mombaur et al.
Li & Todorov
Ziebart
This lecture
Lecture 52 of 73 of Deep Reinforcement Learning.
Choose a node to see how it connects to this lecture. What comes before is on the leftat the top. What comes after is on the rightat the bottom. The ring is what it teaches. People and places are above. Other lectures and books are below.
Press a chosen node again to open it.