Seminar series of the Mathematical Foundations group
Tuesdays 13.15 - 15.15 in 1W 2.102
Seminars are open to all
Willem Heijltjes (University of Bath)
Proof Nets and Complexity
In this talk I will give an overview of some recent and some very recent developments in linear logic proof nets.
John Power (University of Bath)
Category theoretic semantics for theorem proving in logic programming: embracing the laxness
(joint with Ekaterina Komendantskaya)
A propositional logic program P may be identified with a Pf Pf-coalgebra on the set of atomic propositions in the program. The corresponding C(Pf Pf)-coalgebra, where C(Pf Pf) is the cofree comonad on Pf Pf), describes derivations by resolution. Using lax semantics, that correspondence may be extended to a class of first-order logic programs without existential variables. The resulting extension captures proofs by term-matching resolution in logic programming. Refining the lax approach, we further extend it to arbitrary logic programs.
Jim Laird (University of Bath)
Sam Staton (University of Oxford)
Harry Gunn (University of Bath, Masters Student)
Paul Harrenstein (University of Oxford)
Giuseppe Primiero (Middlesex University)
SecureND: Natural Deduction for Secure Trust
Applications in computational domains complement verified knowledge with information sharing processes. From a logical viewpoint, formulating assertion operations in terms of a trust function is challenging, both conceptually and technically. In this talk we overview SecureND, a natural deduction calculus for knowledge derivation under trust. Its design is motivated by the problem of trust transitivity. We present also its implementation as the Coq protocol SecureNDC, to deal with trusted sources in software management systems. We conclude with an overview of current and future extensions of our language.
Andrea Aler Tubella (University of Bath)
A generalised cut-elimination procedure through subatomic logic
Through subatomic logic we are able to present sufficient conditions for a proof system to enjoy cut-elimination. In this talk I will present subatomic logic, how it enables us to present proof systems that have single (linear) rule scheme and a recent result: we can generalise the splitting procedure for cut-elimination to any proof system whose rules and connectives have certain properties.
17 & 24 November
John Power (University of Bath)
I plan to give two talks about Lawvere theories. These are not for experts but rather to give the details. Lawvere introduced the notion in his PhD thesis in 1963, providing not only a category theoretic account of universal algebra but one that is also presentation-independent. Remarkably, his definition was embraced, albeit with a caveat and in different terms, by universal algebraists but not by category theorists. The latter, from 1966, generally preferred to model universal algebra owing to a little more generality but at a very considerable cost. Computer scientists were then influenced to adopt monads, but much could and has been gained by recasting some of the latter's concerns in terms of Lawvere theories. Ultimately, I think Lawvere theories are a superior approach, but benefit very much from the relationship with monads, and I duly plan to explain it.
Guillaume Munch-Maccagnoni (University of Cambridge)
Polarised realizability structures, models, and depolarisation
Polarisation describes the presence of an evaluation order, and is characterised denotationally by a non-associativity of compositions. We recently proposed a polarised, Curry-style approach to the λ-calculus with extensional sums, in correspondence with polarised intuitionistic logic. We suggested that associativity of composition in this context should not be seen as a syntactic axiom, but as an emergent property akin to termination. Traditionally, issues with sums in denotational semantics have rather been considered to be with extensionality than with the associativity. This will be explained in an introductory fashion in a first part.
In a second part, I will more formally relate the termination in the λ-calculus with sums to depolarisation, i.e. associativity of composition, or more familiarly the fact that the order of evaluation does not matter. First, a general setting of polarised realizability structures for polarised calculi with or without control operators is developed. Then, a general technique to build observational models from these structures is explained. Finally, under broad conditions, the observational models that the non-associative syntactic structure gives rise to satisfy the associativity of composition (and are therefore cartesian closed categories with binary co-products). I will sketch an analogy between intuitionistic depolarisation and parametricity.
Matthijs Vákár (University of Oxford)
Game Semantics for Dependent Types
Game semantics can act as a unifying semantic framework, providing compelling models for a strikingly wide range of programming languages, type theories and logics. A notable exception has been dependent type theory, which had so far defied a game theoretic description. We present a proposal to fill this gap in the form of a new categorical model of dependent type theory, based on a category of games and history-free winning strategies. We model dependent type theory with 1-, Sigma-, Pi- and intensional Id-types as well as finite inductive type families (which act as ground types, like calendars). We discuss the place of the Id-types in the intensionality spectrum as well as the strong completeness properties the model satisfies.
Most of the talk should be understandable without prior knowledge of game semantics and dependent type theory.
Ugo dal Lago (Universitá di Bologna)
Higher-Order Probabilistic Computation: Calculi, Observational Equivalence, and Implicit Complexity
Probabilistic models are more and more pervasive in computer science, and randomized algorithms are the ones offering the best performances in many domains. Higher-order probabilistic computation – in which a probabilistic function may be passed as a parameter and returned as a result – is on the other hand a relatively underdeveloped field, which is however receiving more and more attention. We give a survey of what is known about probabilistic lambda-calculi, later focusing on some of our recent results on implicit complexity and on inductive and coinductive techniques for program equivalence. Finally, we hint at how all this could be useful when structuring proofs of security for cryptographic primitives, but also when expressing probabilistic models in the context of machine learning.
Willem Heijltjes (University of Bath)
Complexity Bounds for Sum-Product Logic via Additive Proof Nets and Petri Nets
This is joint work with Dominic Hughes. We investigate efficient algorithms for the additive fragment of linear logic. This logic is an internal language for categories with finite sums and products, and describes concurrent two-player games of finite choice. In the context of session types, typing disciplines for communication along channels, the logic describes the communication of finite choice along a single channel.
We give a simple linear time correctness criterion for unit-free propositional additive proof nets via a natural construction on Petri nets. This is an essential ingredient to linear time complexity of the combinatorial proofs for classical logic by Dominic Hughes.
For full propositional additive linear logic, including the units, we give a proof search algorithm that is linear-time in the product of the source and target formula, and an algorithm for proof net correctness that is of the same time complexity. We prove that proof search in first-order additive linear logic is NP-complete.
Anupam Das (ENS Lyon)
A complete axiomatisation of MSOL on infinite trees.
We show that an adaptation of Peano's axioms for second-order arithmetic to the language of monadic second-order logic (MSOL) completely axiomatises the associated theory (SkS) over infinite trees. This continues a line of work begun by Büchi and Siefkes with axiomatisations of MSOL over various classes of linear orders. Our proof formalises, in the axiomatisation, a translation of MSO formulas to alternating parity tree automata. The main ingredient is the formalised proof of positional determinacy for the corresponding parity games which, as usual, allows us to complement automata and to deal with the negation of MSO formulas. The Comprehension Scheme of MSOL is used to obtain uniform winning strategies, whereas most usual proofs of positional determinacy rely on instances of the Axiom of Choice or transfinite induction. (Consequently we obtain an alternative decision procedure for MSOL over infinite trees, via proof search, that remains entirely internal to the language.)
This talk is based on joint work with Colin Riba that will be presented at LICS '15.
Georg Struth (University of Sheffield)
Completeness Theorems for Bi-Kleene Algebras and Series-Parallel Rational Pomset Languages
Pomsets form a standard model of true concurrency. In this lecture I present a completeness result for a class of pomset languages, which generalises the regular languages to the realm of concurrency. More precisely I show that the congruence on series-parallel rational pomset expressions generated by series-parallel rational pomset language identity is axiomatised by the axioms of Kleene algebra plus those of commutative Kleene algebra. A decision procedure is extracted from this proof. On the way to this result, series-parallel rational pomset languages are proved to be closed under the operations of co-Heyting algebras and homomorphisms. These results form a significant step towards a decision procedure for the equational theory of concurrent Kleene algebras, which have recently been proposed for concurrency verification (joint work with Michael Laurence).
James Brotherston (University College London)
Parametric completeness for separation theories (via hybrid logic).
In this talk, we consider the logical gap between the following two concepts:
(1) provability in a propositional axiomatisation of separation logic, which is usually given by the bunched logic BBI; and
(2) validity in an intended class of models of separation logic, as commonly considered in its program verification applications. Such intended classes are usually specified by a collection of algebraic axioms describing specific model properties, which we call a separation theory.
Here, we show first that several typical properties of separation theories are in fact not definable in BBI. Then, we show that these properties become definable in a natural hybrid extension of BBI, obtained by adding a theory of naming to BBI in the same way that hybrid logic extends normal modal logic. Finally, we show how to build an axiomatic proof system for our hybrid logic in such a way that adding any axiom of a certain form yields a sound and complete proof system with respect to the models satisfying those axioms. In particular, this yields sound and complete proof systems for any separation theory from our considered class (which, to the best of our knowledge, includes all those appearing in the literature).
This is joint work with Jules Villard, now at Facebook.
Guilhem Jaber (Queen Mary)
Reasoning on Equivalence of Stateful Programs with Operational Game Semantics
Contextual equivalence of programs written in a functional language with references (i.e. local mutable states) is a notoriously hard problem, specially with higher-order references (i.e. references which can store functions). In the last twenty years, different techniques have been introduce to that purpose: Kripke Logical Relations, Bisimulations and Algorithmic Game Semantics.
In this talk, we will see how to use operational game semantics, namely the trace semantics for a language with references introduced by Laird, to build a new technique, Kripke Open Bisimulations, to reason on equivalence of programs, taking the best of the previous methods. This technique is simple enough to be mostly automatized: it becomes possible to model-check equivalence of programs.
If time permits, we will see how to extend this technique to polymorphism.