Yuille, Alan L. and Stolorz, Paul and Utans, Joachim (1994) Statistical physics, mixtures of distributions, and the EM algorithm. Neural Computation, 6 (2). pp. 334-340. ISSN 0899-7667. http://resolver.caltech.edu/CaltechAUTHORS:YUInc94
- Published Version
See Usage Policy.
Use this Persistent URL to link to this item: http://resolver.caltech.edu/CaltechAUTHORS:YUInc94
We show that there are strong relationships between approaches to optmization and learning based on statistical physics or mixtures of experts. In particular, the EM algorithm can be interpreted as converging either to a local maximum of the mixtures model or to a saddle point solution to the statistical physics system. An advantage of the statistical physics approach is that it naturally gives rise to a heuristic continuation method, deterministic annealing, for finding good solutions.
|Additional Information:||© 1994 Massachusetts Institute of Technology. Posted Online April 10, 2008. We would like to thank Eric Mjolsness and Anand Rangarajan for helpful conversations and encouragement. One of us (A.L.Y.) thanks DARPA and the Air Force for support under contract F49620-92-J-0466 and Geoffrey Hinton for a helpful conversation.|
|Usage Policy:||No commercial reproduction, distribution, display or performance rights in this work are provided.|
|Deposited By:||Jason Perez|
|Deposited On:||18 Jun 2009 18:20|
|Last Modified:||26 Dec 2012 10:53|
Repository Staff Only: item control page