|
Organizers |
SMV-algebras and Probabilistic Kripke Models
by
Tommaso Flaminio
Dipartimento di Matematica e Informatica, Università degli Studi di Siena. Italy
Coauthors: Franco Montagna
States on MV-algebras have been introduced by Mundici in [Mu] as averaging processes for formulas in \Lukasiewicz logic.
In order to treat states in a logical framework, Flaminio and Godo introduce in [FG07] the logics FP(\Ln, \L) and FP(\L, \L). The latter is obtained by adding a unary modality Pr to the language of \Lukasiewicz logic (cf [H98]) and modal axioms suggested by the following semantic interpretation: the probability of an event A is interpreted as the truth value of the modal formula Pr(A).
Well-founded formulas of FP(\L, \L) are constituted of all the formulas of \Lukasiewicz logic (those are the non-modal formulas), and the class of modal formulas so defined: for each non-modal formula A, Pr(A) is a modal formula, the truth constant [`0] is modal, finally these formulas are combined by means of the \Lukasiewicz connectives. This means that, for instance, neither Pr(A→Pr(B)) nor B⊕Pr(A) (A and B being Pr-free) are well-founded formulas. Using those modal fuzzy logics one can treat probability over many-valued events.
A probabilistic Kripke model for the logic FP(\L, \L) is a pair K=(W, m) where W is a set of valuations of \Lukasiewicz formulas in [0, 1] and m:W→[0, 1] satisfies the condition: ∑w ∈ Wm(w)=1. Elements of W are also called nodes or possible worlds.
Given a Kripke model K=(W, m) and a formula A of FP(\L, \L), the truth value ∥A∥K, w of A in K at the node w is inductively defined as follows:
As they are, the FP(\Ln, \L) and FP(\L, \L) logics are not algebraizable in the sense of Blok-Pigozzi. Recall in fact that Pr(A) is a well-founded formula only if A is a non-modal formula (and hence A does not contain any occurrence of Pr), and a formula of the form B⊕C is well-founded whenever B is modal iff C is modal. Therefore the algebraic counterpart of the operator Pr is a partial operation but not an operation.
As to provide an algebraic approach to states, in [FM07] we introduced the variety of SMV-algebras. An SMV-algebra is a system A=(A, ⊕, ¬, s, 0, 1), where the reduct (A, ⊕, ¬, 0, 1) is an MV-algebra, and s:A→ A satisfies the following equations for each x, y ∈ A:
In [FM07, FM08] we show how, starting from an SMV-algebra A one can define a state s (in the sense of Mundici) on the MV-reduct of A. Furthermore, we use SMV-algebras to equationally characterize the coherence of a finite and rational-valued assessment over \Lukasiewicz events:
Theorem 1 Let a: P(Ai)=ni/mi be a rational assessment over the \Lukasiewicz formulas A1, ..., At. Then the following are equivalent:
Let now FP+(\L, \L) be the logic obtained by extending FP(\L, \L) by the following:
An FP+(\L, \L)-formula A is said 1-satisfiable if there exists a probabilistic Kripke model K such that ∥A∥K=1. Our main result now reads as follows:
Theorem 2 Let A be a term in the language of SMV-algebras. Then the following are equivalent:
Using this result we also show that the problem of deciding whether an equation A=1 holds in an SMV-algebra is PSPACE. The latter result, together with Theorem 1, provides a (new proof (cf [H07]) for the) PSPACE-containment for the problem of establishing the coherence of a rational assessment over \Lukasiewicz formulas.
Date received: February 21, 2008
Copyright © 2008 by the author(s). The author(s) of this document and the organizers of the conference have granted their consent to include this abstract in Atlas Conferences Inc. Document # cavs-13.