BIB-VERSION:: CS-TR-v2.0 ID:: SBCS//stark/bisimulation.ps.gz ENTRY:: January 3, 2003 ORGANIZATION:: State University of New York at Stony Brook, Computer Science TITLE:: Implementation of a Compositional Performance Analysis Algorithm for Probabilistic I/O Automata TYPE:: Technical Report AUTHOR:: Stark, Eugene W. CONTACT:: Eugene W. Stark, Department of Computer Science, SUNY at Stony Brook, Stony Brook, NY 11794-4400 Tel: 631-632-8444 DATE:: January 3, 2003 RETRIEVAL:: HTTP from BSD7.CS.SUNYSB.EDU with the URL http://bsd7.cs.sunysb.edu/~pioa/Papers/bisimulation.abstract ABSTRACT:: Previous work of the author has developed "probabilistic input/output automata" (PIOA) as a formalism for modeling systems that exhibit concurrent and probabilistic behavior. Central to that work was the notion of the "behavior map" associated with a state of a PIOA. The present paper presents a new, simpler definition for PIOA behavior maps, investigates the induced "same behavior map" equivalence relation, and compares it with the standard notion of probabilistic bisimulation equivalence. "Weighted finite automata" are used as a unifying formalism to facilitate the comparison. A general notion of congruence for weighted automata is defined, which relates "signed measures" on states, rather than just individual states. PIOA are defined as a class of weighted automata, as are the class of "probabilistic weighted automata" for which the standard definition of probabilistic bisimulation makes sense. A characterization is obtained of probabilistic bisimulation as the largest congruence that is in a sense generated by its restriction to a relation on states. This characterization is then used as the definition of "weighted bisimulation", which generalizes probabilistic bisimulation to the full class of weighted automata. PIOA behavior equivalence is also shown to define a weighted automata congruence, which is strictly refined by weighted bisimulation equivalence. The relationship between these congruences and a notion of "composition" for weighted automata is also examined.