Obtaining measure concentration from Markov contraction
نویسنده
چکیده
Concentration bounds for non-product, non-Haar measures are fairly recent: the first such result was obtained for contracting Markov chains by Marton in 1996. Since then, several other such results have been proved; with few exceptions, these rely on coupling techniques. Though coupling is of unquestionable utility as a theoretical tool, it appears to have some limitations. Coupling has yet to be used to obtain bounds for more general Markov-type processes: hidden (or partially observed) Markov chains, Markov trees, etc. As an alternative to coupling, we apply the elementary Markov contraction lemma in a novel way, to obtain simple, useful, and apparently new concentration results for the various Markov-type processes. Our technique is generic and holds the potential to yield numerous results in this vein.
منابع مشابه
Measure Concentration of Markov Tree Processes
We prove an apparently novel concentration of measure result for Markov tree processes. The bound we derive reduces to the known bounds for Markov processes when the tree is a chain, thus strictly generalizing the known Markov process concentration results. We employ several techniques of potential independent interest, especially for obtaining similar results for more general directed acyclic ...
متن کاملMeasure Concentration of Hidden Markov Processes
We prove what appears to be the first concentration of measure result for hidden Markov processes. Our bound is stated in terms of the contraction coefficients of the underlying Markov process, and strictly generalizes the Markov process concentration results of Marton (1996) and Samson (2000). Somewhat surprisingly, the hidden Markov process is at least as “concentrated” as its underlying Mark...
متن کاملMarkov extensions and conditionally invariant measures for certain logistic maps with small holes
We study the family of quadratic maps fa(x) = 1 − ax 2 on the interval [−1, 1] with 0 ≤ a ≤ 2. When small holes are introduced into the system, we prove the existence of an absolutely continuous conditionally invariant measure using the method of Markov extensions. The measure has a density which is bounded away from zero and is analogous to the density for the corresponding closed system. Thes...
متن کاملConcentration of Measure and Mixing for Markov Chains
We consider Markovian models on graphs with local dynamics. We show that, under suitable conditions, such Markov chains exhibit both rapid convergence to equilibrium and strong concentration of measure in the stationary distribution. We illustrate our results with applications to some known chains from computer science and statistical mechanics.
متن کاملUtterance Verification Using State-Level Log-Likelihood Ratio with Frame and State Selection
This paper suggests utterance verification system using state-level log-likelihood ratio with frame and state selection. We use hidden Markov models for speech recognition and utterance verification as acoustic models and anti-phone models. The hidden Markov models have three states and each state represents different characteristics of a phone. Thus we propose an algorithm to compute state-lev...
متن کامل