The popular Science On Screen series returns to Santa Fe Wednesday evening, May 8, with Simon DeDeo and the 1992 cult hacker film Sneakers.
SFI's 2013 Community Lecture series debuted March 14 with UC-Boulder's Leysia Palen describing how victims, observers, and “citizen-responders” are using modern technology to participate in disaster response. Watch ...
Speaking at SFI yesterday, noted climate scientist James Hansen told an overflow crowd that efforts to stem climate change will be ineffectual as long as fossil fuels remain the cheapest ...
SFI's crowdfunding campaign has reached its goal. The resulting research will help scientists preserve the threatened landscapes on which indigenous human groups depend.
The 2012 Science On Screen series in Santa Fe wrapped up December 13 to a full house, with "The Gods Must Be Crazy" and Murray Gell-Mann's distinctive insight and ...
Seminar
December 06, 2012
12:15 PM
Collins Conference Room
Aram Galstyan (University of Southern California)
Abstract. A Hidden Markov Model (HMM) is a Markov process observed through a noisy channel. HMMs are used extensively for modeling sequential data and have various applications in natural language processing, bioinformatics, mathematical economics, and so on. One of the main problems underlying HMMs is inferring the hidden state sequence based on noise-corrupted observations, which is usually solved via the maximum a posteriori (MAP) method. Here I present a statistical physics analysis of MAP estimation for a simple binary HMM, by reducing it to an appropriately defined Ising model. It is shown that for sufficiently low noise levels, the MAP solution is uniquely related to the observed sequence. Above a critical noise level, however, there are exponentially many solutions to the estimation problem, which means that one cannot uniquely reconstruct the hidden state sequence based on observations. I will also discuss how one can mitigate this undesirable feature by active inference, i.e., by adaptively acquiring information about the (true) hidden states.
Purpose: Research Collaboration
SFI Host: Cris Moore
Facebook
Twitter
LinkedIn
Email Newsletter
SFI News Feed
Google+
YouTube
iTunes U