State Grammars with Stores - Descriptional Complexity of Formal Systems Access content directly
Conference Papers Year : 2018

State Grammars with Stores

Oscar H. Ibarra
  • Function : Author
  • PersonId : 1038032
Ian Mcquillan
  • Function : Author
  • PersonId : 1038033

Abstract

State grammars are context-free grammars where the productions have states associated with them, and can only be applied to a nonterminal if the current state matches the state in the production. Once states are added to grammars, it is natural to add various stores, similar to machine models. With such extensions, productions can only be applied if both the state and the value read from each store matches between the current sentential form and the production. Here, generative capacity results are presented for different derivation modes, with and without additional stores. In particular, with the standard derivation relation, it is shown that adding reversal-bounded counters does not increase the capacity, and states are enough. Also, state grammars with reversal-bounded counters that operate using leftmost derivations are shown to coincide with languages accepted by one-way machines with a pushdown and reversal-bounded counters, and these are surprisingly shown to be strictly weaker than state grammars with the standard derivation relation (and no counters). Complexity results of some decision problems involving state grammars with counters are also studied.
Fichier principal
Vignette du fichier
470153_1_En_14_Chapter.pdf (258.64 Ko) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

hal-01905629 , version 1 (26-10-2018)

Licence

Attribution

Identifiers

Cite

Oscar H. Ibarra, Ian Mcquillan. State Grammars with Stores. 20th International Conference on Descriptional Complexity of Formal Systems (DCFS), Jul 2018, Halifax, NS, Canada. pp.163-174, ⟨10.1007/978-3-319-94631-3_14⟩. ⟨hal-01905629⟩
26 View
64 Download

Altmetric

Share

Gmail Facebook X LinkedIn More