EN FR
EN FR


Section: New Results

Stochastic mechanics of graph rewriting

Participants : Nicolas Behr, Vincent Danos, Ilias Garnier [correspondant] .

We propose an algebraic approach to stochastic graph-rewriting which extends the classical construction of the Heisenberg-Weyl algebra and its canonical representation on the Fock space. Rules are seen as particular elements of an algebra of “diagrams": the diagram algebra D. Diagrams can be thought of as formal computational traces represented in partial time. They can be evaluated to normal diagrams (each corresponding to a rule) and generate an associative unital non-commutative algebra of rules: the rule algebra R. Evaluation becomes a morphism of unital associative algebras which maps general diagrams in D to normal ones in R. In this algebraic reformulation, usual distinctions between graph observables (real-valued maps on the set of graphs defined by counting subgraphs) and rules disappear. Instead, natural algebraic substructures of R arise: formal observables are seen as rules with equal left and right hand sides and form a commutative subalgebra, the ones counting subgraphs forming a sub-subalgebra of identity rules. Actual graph-rewriting is recovered as a canonical representation of the rule algebra as linear operators over the vector space generated by (isomorphism classes of) finite graphs. The construction of the representation is in close analogy with and subsumes the classical (multi-type bosonic) Fock space representation of the Heisenberg-Weyl algebra.

This shift of point of view, away from its canonical representation to the rule algebra itself, has unexpected consequences. We find that natural variants of the evaluation morphism map give rise to concepts of graph transformations hitherto not considered. These will be described in a separate paper [2]. In this extended abstract we limit ourselves to the simplest concept of double-pushout rewriting (DPO). We establish “jump-closure", i.e. that the sub-space of representations of formal graph observables is closed under the action of any rule set. It follows that for any rule set, one can derive a formal and self-consistent Kolmogorov backward equation for (representations of) formal observables.

This result and the fallowing ones, co-authored by Vincent Danos, were published in peer-reviewed international conferences and journals. Although the papers are on HAL, they are not imported in the bibtex file so we can't cite them properly.