Skip to Main content Skip to Navigation

Infinite Computations in Algorithmic Randomness and Reverse Mathematics

Abstract : This thesis focuses on the gains of infinite time computations to mathematical logic. Infinite time computations is a variant of the traditional definition of computations as a finite sequence of stages, each stage being defined from the previous ones, and finally reaching a halting state. In this thesis, we consider the case where the number of stages is not necessarily finite, but can continue along ordinals, an extension of the integers. There exists several ways to implement this idea, we will use three of them: higher recursion, infinite time Turing machines and α-recursion.Part of this works concerns the domain of reverse mathematics, and especially Hindman's theorem. Reverse mathematics is a program consisting in the study of theorems and axioms from the point of view of their "strength", and establishing a hierarchy on these. In particular the question of which axioms are needed in a proof of a given statement is central. We study Hindman's theorem under this lens, a combinatorial result from Ramsey's theory stating that for every partitioning of the integers into finitely many colors, there must exists an infinite set such that any sum of elements taken from it has a fixed color. In this thesis, we make some progress in the question of the minimal axiomatic system needed to show this result, by showing that the existence of some intermediate combinatorial objects is provable in a weak system.Weihrauch reduction is a way to compare the strength of theorems, that has been introduced in reverse mathematics recently. It sees theorems as problems to solve, and then compare their difficulties. This reduction is still less studied in this context, in particular few of the most important principles of reverse mathematics are not yet well comprehended. One of these is the Arithmetical Transfinite Recursion principle, an axiomatic system with strong links with infinite time computations and especially higher recursion. We continue the study of this principle by showing its links with a particular type of axiom of choice, and use it to separate the dependent and independent version of this choice.Yet another field of mathematical logic that benefits from computability theory is the one of algorithmic randomness. It studies "random" reals, those that it would seem reasonable to think that they arise from a process picking a real uniformly in some interval. A way to study this is to considerate, for a given real, the smallest algorithmic complexity of a null set containing it. This domain has proven very rich and has already been extended to certain type of infinite time computation, thereby modifying the complexity class considered. However, it has been extended to infinite time Turing machine and α-recursion only recently, by Carl and Schlicht. In this thesis, we contribute to the study of the most natural randomness classes for ITTMs and α-recursion. We show that two important classes, Σ-randomness and ITTM-randomness, are not automatically different; in particular their categorical equivalent are in fact the same classes.
Document type :
Complete list of metadata

Cited literature [71 references]  Display  Hide  Download
Contributor : Abes Star :  Contact
Submitted on : Tuesday, September 15, 2020 - 4:20:09 PM
Last modification on : Wednesday, September 16, 2020 - 3:27:05 AM


Version validated by the jury (STAR)


  • HAL Id : tel-02936757, version 2



Paul-Elliot Anglès d'Auriac. Infinite Computations in Algorithmic Randomness and Reverse Mathematics. Logic in Computer Science [cs.LO]. Université Paris-Est, 2019. English. ⟨NNT : 2019PESC0061⟩. ⟨tel-02936757v2⟩



Record views


Files downloads