2001 ATutorialOnVariatApproxMethods

• (Jaakkola, 2001) ⇒ Tommi S. Jaakkola. (2001). “Tutorial on Variational Approximation Methods.” In: (Opper & Saad, 2001)

Subject Headings:

Cited By

~158 http://scholar.google.com/scholar?cites=12062150500178786044

Quotes

Abstract

• We provide and introduction to the theory and use of variational methods for inference and estimation in the context of graphical models. Variational methods become useful as efficient approximate methods when the structure of the graph model no longer admits feasible exact probabilistic calculation. The emphasis of this tutorial is on illustrating how inference and estimation problems can be transfered into variational form along with describing the resulting approximation algorithms and their properties insofar as these are currently known.
• http://people.csail.mit.edu/tommi/papers/Jaa-nips00-tutorial.pdf
  • Let P(x1,...,xn) be the distribution of interest over n variables
  • We divide the set of variables into
  • 1. “visible" variables xv whose marginal distribution P(xv) we are interested in computing
  • 2. “hidden" variables xh whose posterior distribution P(xh|xv) we want
  • Evaluating the marginal or posterior involves summing over all con figurations of the hidden variables xh ⇒ P=...
  • The complexity of this operation depends on the structure or factorization of the joint distribution P
  • We try to capture the factorization explicitly in terms of graphs

References

---

,

	Author	volume	Date Value	title	type	journal	titleUrl	doi	note	year
2001 ATutorialOnVariatApproxMethods	Tommi S. Jaakkola			Tutorial on Variational Approximation Methods			http://lcib.rutgers.edu/~james/variational methods.pdf			2001