1: \begin{abstract}

2: %--------------------------------------------------------------

3: Motivated by the Markov chain Monte Carlo (MCMC) approach to the compression

4: of discrete sources developed by Jalali and Weissman, 

5: we propose a lossy compression algorithm for analog

6: sources that relies on a finite reproduction alphabet, which

7: grows with the input length. The algorithm achieves, in an appropriate

8: asymptotic sense, the optimum Shannon theoretic tradeoff between 

9: rate and distortion, universally for stationary ergodic

10: continuous amplitude sources.

11: We further  propose an MCMC-based algorithm that resorts to a reduced

12: reproduction alphabet when such reduction does not prevent

13: achieving the Shannon limit. The latter algorithm is 

14: advantageous due to its reduced complexity and improved rates of 

15: convergence when employed on sources with a finite and small

16: optimum reproduction alphabet.

17: %--------------------------------------------------------------

18: \end{abstract}

19: