Title :
Extension of the PAC framework to finite and countable Markov chains
Author_Institution :
IBM T. J. Watson Res. Center, Yorktown Heights, NY, USA
fDate :
1/1/2003 12:00:00 AM
Abstract :
We consider a model of learning in which the successive observations follow a certain Markov chain. The observations are labeled according to a membership to some unknown target set. For a Markov chain with finitely many states we show that, if the target set belongs to a family of sets with a finite Vapnik-Chervonenkis (1995) dimension, then probably approximately correct (PAC) learning of this set is possible with polynomially large samples. Specifically for observations following a random walk with a state space 𝒳 and uniform stationary distribution, the sample size required is no more than Ω(t0/1-λ2log(t0|χ|1/δ)), where δ is the confidence level, λ2 is the second largest eigenvalue of the transition matrix, and t0 is the sample size sufficient for learning from independent and identically distributed (i.i.d.) observations. We then obtain similar results for Markov chains with countably many states using Lyapunov function technique and results on mixing properties of infinite state Markov chains.
Keywords :
Lyapunov methods; Markov processes; learning systems; probability; set theory; state-space methods; Lyapunov function; PAC framework extension; confidence level; eigenvalue; finite Markov chains; finite Vapnik-Chervonenkis dimension; i.i.d. observations; independent identically distributed observations; learning model; mixing properties; polynomially large samples; probably approximately correct model; random walk; set membership; state space; successive observations; transition matrix; uniform stationary distribution; Eigenvalues and eigenfunctions; Lyapunov method; Neural networks; Polynomials; State-space methods; Steady-state; Stochastic processes; Virtual colonoscopy;
Journal_Title :
Information Theory, IEEE Transactions on
DOI :
10.1109/TIT.2002.806131