new
  • (with George Yin) Analysis of LMS Algorithm for Slowly Varying Markovian Parameters -- Tracking Slow Hidden Markov Models and Adaptive Multiuser Detection in DS/CDMA,
  • This paper gives an example where iterate averaging can result in a tracking stochastic approximation (i.e. constant step size) algorithm with optimal convergence rate. Previously iterate averaging has been shown to result in optimal asymptotic convergence rate for decreasing step size algorithms. To our knowledge, this is the first time an example has been given of a case where iterate averaging can lead to a tracking algorithm with asymptotically optimal convergence rate! We also show how to utilize these result to analyse the effect of Markovian type admission and access policies on an adaptive multiuser deteciton algorithm.

  • V. Krishnamurthy, Algorithms for Optimal Scheduling of Hidden Markov Model Sensors, IEEE Trans Signal Processing, Vol.50, No.6, pp.1382--1397, June 2002.
  • V. Krishnamurthy and G. Yin, Recursive Algorithms for Estimation of Hidden Markov Models and Autoregressive Models with Markov Regime , IEEE Transactions on Information Theory, Vol.48, No.2, pp.458--476, Feb 2002. 

  • vikram krishnamurthy
    Last modified: Thu Jun 5 09:46:28 PDT 2003