MOVES Seminar 10 Mar, 2011, 10:00

Weighted Lumpability on Markov Chains.

Abstract - This talk reconsiders Bernardo's T-lumpability on continuous-time Markov chains (CTMCs). This notion allows for a more aggressive state-level aggregation than ordinary lumpability. We provide a novel structural definition of (what we refer to as) weighted lumpability, prove some elementary properties, and investigate its compatibility with linear real-time objectives.
The main result is that the probability of satisfying a deterministic timed automaton specification coincides for a CTMC and its weigthed lumped analogue. The same holds for metric temporal logic formulas.