On the Power and Universality of Biologically-inspired Models of Computation

Abstract : The present thesis considers the problems of computational completeness and universality for several biologically-inspired models of computation: insertion-deletion systems, networks of evolutionary processors, and multiset rewriting systems. The presented results fall into two major categories: study of expressive power of the operations of insertion and deletion with and without control, and construction of universal multiset rewriting systems of low descriptional complexity. Insertion and deletion operations consist in adding or removing a subword from a given string if this subword is surrounded by some given contexts. The motivation for studying these operations comes from biology, as well as from linguistics and the theory of formal languages. In the first part of the present work we focus on insertion-deletion systems closely related to RNA editing, which essentially consists in inserting or deleting fragments of RNA molecules. An important feature of RNA editing is the fact that the locus the operations are carried at is determined by certain sequences of nucleotides, which are always situated to the same side of the editing site. In terms of formal insertion and deletion, this phenomenon is modelled by rules which can only check their context on one side and not on the other. We show that allowing one-symbol insertion and deletion rules to check a two-symbol left context enables them to generate all regular languages. Moreover, we prove that allowing longer insertion and deletion contexts does not increase the computational power. We further consider insertion-deletion systems with additional control over rule applications and show that the computational completeness can be achieved by systems with very small rules. The motivation for studying insertion-deletion systems also comes from the domain of computer security, for the purposes of which a special kind of insertion-deletion systems called leftist grammars was introduced. In this work we propose a novel graphical instrument for visual analysis of the dynamics of such systems. The second part of the present thesis is concerned with the universality problem, which consists in finding a fixed element able to simulate the work any other computing device. We start by considering networks of evolutionary processors (NEPs), a computational model inspired by the way genetic information is processed in the living cell, and construct universal NEPs with very few rules. We then focus on multiset rewriting systems, which model the chemical processes running in the biological cell. For historical reasons, we formulate our results in terms of Petri nets. We construct a series of universal Petri nets and give several techniques for reducing the numbers of places, transitions, inhibitor arcs, and the maximal transition degree. Some of these techniques rely on a generalisation of conventional register machines, proposed in this thesis, which allows multiple register checks and operations to be performed in a single state transition
Document type :
Theses
Complete list of metadatas

Cited literature [119 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-01272318
Contributor : Abes Star <>
Submitted on : Wednesday, February 10, 2016 - 3:45:07 PM
Last modification on : Wednesday, December 19, 2018 - 3:50:03 PM
Long-term archiving on : Saturday, November 12, 2016 - 4:54:45 PM

File

TH2015PEST1012.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-01272318, version 1

Collections

Citation

Sergiu Ivanov. On the Power and Universality of Biologically-inspired Models of Computation. Computation and Language [cs.CL]. Université Paris-Est, 2015. English. ⟨NNT : 2015PESC1012⟩. ⟨tel-01272318⟩

Share

Metrics

Record views

265

Files downloads

229