Online Worldwide Seminar on
Logic and Semantics (OWLS)
The Online Worldwide Seminar on Logic and Semantics is a new series of research talks, highlighting the most exciting recent work in the international computer science logic community. The scope of the seminar series is roughly that of the major computer science logic conferences such as LICS, ICALP and FSCD. It takes place on most Wednesdays, with a focus every other week on the work of young researchers.
In this time of restricted international travel, a key aim of this series is to provide a forum for the informal discussion and social interaction that is so important for the progress of science. To facilitate this, the seminar incorporates in virtual form a number of features more normally associated with physical meetings:
Talks are given by invitation of the organizing committee. Feedback is warmly encouraged; if you have any comments or ideas, feel free to get in touch with an organizer.
The OWLS mailing list sends reminders before each seminar, and announces upcoming seminars. To subscribe, or manage an existing subscription, visit this link:
Messages will be sent from the address email@example.com.
Seminars take place on Wednesdays at 2pm UTC. An easy way to see the corresponding time in your own time zone is to add the OWLS calendar to your own calendar, with the following links: Google Calendar Apple Calendar
Seminars are 45 minutes long including questions, and will start at the advertised time. As described below, feel free to join early, or stay around after the end of the talk, to participate in the coffee breaks. A link to join the seminar is given in the participant information below. Every other week, the seminar hosts a young researcher to present their work; these seminars are labelled OWLS-YR.
Here is the list of upcoming seminars.
While model checking has often been considered as a practical alternative to building formal proofs, we argue here that the theory of sequent calculus proofs can be used to provide an appealing foundation for model checking. Since the emphasis of model checking is on establishing the truth of a property in a model, we rely on the proof-theoretic notion of additive inference rules, since such rules allow provability to directly describe truth conditions. Unfortunately, the additive treatment of quantifiers requires inference rules to have infinite sets of premises and the additive treatment of model descriptions provides no natural notion of state exploration. By employing a focused proof system, it is possible to construct large-scale, synthetic rules that also qualify as additive but contain elements of multiplicative inference. These additive synthetic rules -- essentially rules built from the description of a model -- allow direct treatment of state exploration. This proof-theoretic framework provides a natural treatment of reachability and non-reachability problems, as well as tabled-deduction, bisimulation, and winning strategies.
This talk is based on a paper co-authored with Quentin Heath appearing in the Journal of Automated Reasoning, 2019.
Here is the list of past seminars in reverse chronological order.
(based on joint works with Alen Arslanagić, Dimitris Kouzapas, Erik Voogd, and Nobuko Yoshida)
Session types are a type-based approach for ensuring correct message-passing programs. A session type specifies the protocol that each channel endpoint in aprogram should respect.
Session types have been originally developed for processes in the pi-calculus, the paradigmatic calculus of interaction and concurrency. While there is substantial value in studying session types for the pi-calculus, it is also insightful to investigate them for programming calculi with a more direct link with (functional) programming languages. To this end, we have studied a core calculus for higher-order concurrency and sessions, called HO. HO is a compact blend of sessions and higher-order concurrency, in which the values exchanged by processes can only be abstractions (functions from names to processes).
In this talk, I will overview two results on the expressivity of HO processes with session types. They show how session types not only delineate and enable encodings; they inform the techniques required to reason about their correctness properties. First, HO and the session pi-calculus are equally expressible: one language can be encoded into the other, up to tight behavioral equivalences. Second, there is a class of so-called minimal session types, which only covers protocols without sequential composition but is still expressive enough to encode all typable HO processes.
We present quantitative separation logic (QSL). In contrast to classical separation logic, QSL employs quantities which evaluate to real numbers instead of predicates which evaluate to Boolean values. The connectives of classical separation logic, separating conjunction and separating implication, are lifted from predicates to quantities. This extension is conservative: Both connectives are backward compatible to their classical analogs and obey the same laws, e.g. modus ponens, adjointness, etc. Furthermore, we present a weakest precondition calculus for quantitative reasoning about probabilistic pointer programs in QSL. This calculus is a conservative extension of both Ishtiaq's, O'Hearn's and Reynolds' separation logic for heap-manipulating programs and Kozen's / McIver and Morgan's weakest preexpectations for probabilistic programs. Our calculus preserves O'Hearn's frame rule, which enables local reasoning - a key principle of separation logic. Finally, we briefly touch upon open questions regarding lower bounds on weakest preexpectations and intensional completeness of our calculus.
Focusing is a general technique that was designed to improve proof search, but has since become increasingly relevant in structural proof theory. The essential idea is to identify and merge the non-deterministic choices in a proof. A focused proof is then given by the alternation of phases where invertible rules are applied eagerly (bottom-up) and phases where the other rules are confined and controlled. The focusing theorem, which asserts the completeness of focused proofs, delivers strong representational benefits. For instance, by ignoring the internal structure of the phases one obtains a well-behaved notion of ‘synthetic rule’ (e.g. they commute with cut-reduction steps). In this talk, we will present a careful study of a family of synthetic rules, the bipoles, giving a fresh view to an old problem: how to incorporate inference rules corresponding to axioms into proof systems for classical and intuitionistic logics. As different synthetic inference rules can be produced for the same axiom, we in fact unify and generalise previous approaches for transforming axioms into rules. This is joint work with Dale Miller, Elaine Pimentel, and Marco Volpe.
One way of thinking of a “pure name” (or “guid") is as a random number, and so names are related to probability. It’s important to understand how some names can be private or secret, and this turns out to correspond to probabilistic concepts. I will talk about semantic models for higher-order programming with probabilities and for programming with names. I’ll show a model of probability that is also a good model of programming with names: it's fully abstract at first order.
This talk will take us past the nu calculus, quasi Borel spaces, some bits of descriptive set theory and some non-parametric statistics. But it will be an introductory talk and I won’t assume familiarity with any of this. The talk will be partly based on our paper at POPL 2021, https://doi.org/10.1145/3434292, which is joint work with Marcin Sabok, Dario Stein and Michael Wolman.
While causal consistency is one of the most fundamental consistency models weaker than sequential consistency, the decidability of safety verification for (finite-state) concurrent programs running under causally consistent shared-memories is still unclear. We establish the decidability of this problem for two standard and well-studied variants of causal consistency. To do so, for each of the variants, we develop an equivalent "lossy" operational semantics, and show that it constitutes a well-structured transition system, which enables decidable verification. The two novel semantics are based on similar key observations, which, we believe, may also be of independent use in the investigation of weakly consistent shared memory models and their verification. Interestingly, our results are in contrast to the undecidability of this problem under the Release/Acquire fragment of the C/C++11 memory model, which forms another variant of a causally consistent memory that, in terms of allowed outcomes, lies strictly between the two models we study. Nevertheless, all these variants coincide for write/write-race-free programs, which implies the decidability of verification for such programs under Release/Acquire.
(Joint work with Udi Boker, partly presented at PLDI'20)
Church's synthesis problem asks whether a synchronous specification from infinite words to infinite words can be realized by a synchronous function. We relax this requirement and ask whether a synchronous specification from infinite words to infinite words can be realized by a computable function, in other words, we simply ask whether a specification is implementable.
This question has been previously studied by Holtmann et al. who showed that the problem is decidable (it was later shown to be EXPtime-complete) for synchronous specifications with total domain. We show that the question is EXPtime-complete for synchronous specifications with partial domain. A specification with partial domain is implementable if for every input sequence in its domain an output sequence can be computed that is correct w.r.t. the specification. A fundamental difference between the total and the partial domain setting is that, if the specification is implementable, in the former setting sequential transducers suffice to do so and in the latter setting a more powerful computation model is needed. We show that if a synchronous specification from infinite words to infinite words is implementable then a deterministic two-way transducer can compute an implementation.
This is joint work with Emmanuel Filiot.
In both computer science and economics, efficiency is a cherished property. In computer science, the field of algorithms is almost solely focused on their efficiency. In economics, the main advantage of the free market is that it promises "economic efficiency." A major lesson from COVID-19 is that both fields have over-emphasized efficiency and under-emphasized resilience. I argue that resilience is a more important property than efficiency and discuss how the two fields can broaden their focus to make resilience a primary consideration. I will include a technical example, showing how we can shift the focus in strategic reasoning from efficiency to resilience.
Iris is a framework for higher-order concurrent separation logic, implemented in the Coq proof assistant, which we have been developing since 2014. Originally designed for pedagogical purposes, Iris has grown into a ongoing, multi-institution project, with active collaborators at Aarhus University, BedRock Systems, Boston College, CNRS/LRI, Groningen University, INRIA, ITU Copenhagen, KU Leuven, Microsoft Research, MIT, MPI-SWS, NYU, Radboud University Nijmegen, Saarland University, and Vrije Universiteit Brussel, and over 35 published papers studying or deploying Iris for verification of complex programs and programming language meta-theory in Rust, Go, OCaml, Scala, and more (https://iris-project.org).
In this talk, we will present two brand new -- and very different -- developments that have the potential to extend the reach of Iris even further. The first is a new *ownership-based refinement type system* for C, which supports *automated* verification of C programs while at the same time being *foundational* (producing Iris proofs in Coq). The second is a complete "remodeling" of Iris, replacing its original step-indexed model with a *transfinite* step-indexed model in order to make Iris suitable for verification of liveness properties.
For this talk, we will not assume any prior knowledge of Iris. Rather, we will briefly review the distinguishing features of Iris, and then explain the key insights behind the aforementioned new developments -- and the problems they are solving -- at a high level of abstraction.
The first new development is joint work with Michael Sammler, Rodolphe Lepigre, Robbert Krebbers, Kayvan Memarian, and Deepak Garg. The second is joint work with Simon Spies, Lennard Gäher, Daniel Gratzer, Joseph Tassarotti, Robbert Krebbers, and Lars Birkedal.
During my PhD research I have studied distributive laws for monads, and developed various no-go theorems. These theorems prove that monads with certain algebraic properties cannot be composed with each other via a distributive law. In this talk I will focus on examples, and share the intuition that I have developed from my results. All the examples in this talk will be based on the Boom hierarchy, a hierarchy of data structures which lends itself perfectly for the study of monad compositions. Based on which monads in this hierarchy do and do not compose with each other via a distributive law, I will make predictions about the behaviour of monad compositions in general.
Mainstream formal analyses of software systems focus on extensional properties of program behaviour. Such properties ultimately deal with the input/output semantics of programs, this way giving information on what programs compute. There are several situations, however, where intensional properties of programs turn out to be crucial to ensure the correct behaviour of software systems. In these scenarios, one needs not only to know what output a program computes, but also how the program computes it. Typical examples of intensional aspects of computations include efficiency, resource usage, robustness with respect to errors, and privacy and security.
In the last decades, these properties have been studied in isolation by means of type systems, denotational semantics, and operational techniques. More recently, uniform accounts of a large family of intensional aspects of computations (oftentimes referred to as coeffectful computations) have been proposed in terms of graded modal types and graded (co)monads.
In this talk, I will present a general framework for studying operationally-based notions of program equivalence for languages with graded modal types. These notions of equivalence identify programs with the same intensional behaviour and thus allow for the intensional analysis of programs. The framework will be instantiated to study the intensional refinement of Abramsky's applicative bisimilarity (a coinductively-defined notion of equivalence for higher-order sequential languages) giving an abstract compositionality theorem generalising well-known results on intensional program behaviour such as Abadi's et al. Non-interference, Pfenning's Proof Irrelevance, and Reed and Pierce's Metric Preservation.
Finally, a formal connection between the theory of intensional program equivalence and the theory of abstract program distance built on top of Lawvere's analysis of generalised metric spaces will be established, this way allowing for the transfer of results and techniques developed in the context of program metric to the realm of intensional analyses of languages with modal types.
Holonomic techniques have deep roots going back to Wallis, Euler, and Gauss, and have evolved in modern times as an important subfield of computer algebra, thanks in large part to the work of Zeilberger and others over the past three decades. In this talk, I will give an overview of the area, and in particular will present a select survey of known and original results on decision problems for holonomic sequences and functions. I will also discuss some surprising connections to the theory of periods and exponential periods, which are classical objects of study in algebraic geometry and number theory; in particular, I will relate the decidability of certain decision problems for holonomic sequences to deep conjectures about periods and exponential periods, notably those due to Kontsevich and Zagier.
Parts of this talk will be based on the paper "On Positivity and Minimality for Second-Order Holonomic Sequences", https://arxiv.org/abs/2007.12282.
It is well known that the classic Łoś-Tarski preservation theorem fails in the finite: there are first-order definable classes of finite structures closed under extensions which are not definable (in the finite) in the existential fragment of first-order logic. We strengthen this by constructing for every n, first-order definable classes of finite structures closed under extensions which are not definable with n quantifier alternations. The classes we construct are definable in the extension of Datalog with negation and indeed in the existential fragment of transitive-closure logic. This answers negatively an open question posed by Rosen and Weinstein.
The talk surveys a series of works that lift the rich semantics and structure of graphs, and the experience of the formal-verification community in reasoning about them, to classical graph-theoretical problems.
Partially observable Markov Decision Processes (POMDPs) are a prime model in sequential decision making. They are heavily studied in operations research, artificial intelligence and verification, to name a few. For POMDPs, a good policy reaches the target with probability at some threshold, and makes its decisions solely based on the previously made observations. Deciding whether a good policy exists is undecidable. One of the methods to solve instances of this reachability problem is to restrict the memory that the strategy may use. This approach is commonly taking, e.g., in reinforcement learning for POMDPs. In the first part of the talk, we consider memoryless strategies in POMDPs. and show that this problem is as hard as the feasibility problem in parametric Markov Chain (pMC) analysis.
In the second part of the talk, we consider this feasibility problem for pMCs. Roughly, a pMC is a Markov chain with symbolic transitions. The feasibility problem asks: Do values for the symbols exist such that in the induced parameter-free Markov chain, a target state is reached with probability at least a half. We discuss the complexity landscape for variants of this decision problem. In particular, we establish that feasibility in pMCs is complete for the complexity class "existential theory of the reals” (ETR). Another example of an ETR-complete problem is deciding whether a multivariate polynomial has a real root. Together with the results of the first half of the talk, this establishes that deciding whether there exists a good memoryless policy in a POMDP is ETR-complete.
This is joint work with Alex Keizer and Jorge A. Pérez.
Compositional methods are central to the development and verification of software systems. They allow to break down large systems into smaller components, while enabling reasoning about the behaviour of the composed system. For concurrent and communicating systems, compositional techniques based on *behavioural type systems* have received much attention. By abstracting communication protocols as types, these type systems can statically check that programs interact with channels according to a certain protocol, whether the intended messages are exchanged in a certain order. For this talk, we will put on our coalgebraic spectacles to investigate *session types*, a widely studied class of behavioral type systems. We will seek a syntax-free description of session-based concurrency as states of coalgebras. The result will be a description of type equivalence, duality, and subtyping relations in terms of canonical coinductive presentations. In turn, this coinductive presentation makes it possible to elegantly derive a decidable type system with subtyping for π-calculus processes, in which the states of a coalgebra will serve as channel protocols. Going full circle, we will also exhibit a coalgebra structure on an existing session type system, and show that the relations and type system resulting from our coalgebraic perspective agree with the existing ones.
Joint work with Danel Ahman, University of Ljubljana.
In modern languages, computational effects are often structured using monads or algebraic effects and handlers. These mechanisms excel at implementation of computational effects within the language itself. For instance, the familiar implementation of mutable state in terms of state-passing functions requires no native state, and can be implemented either as a monad or using handlers. One is naturally drawn to using these techniques also for dealing with actual effects, such as manipulation of native memory and access to hardware. These are represented inside the language as algebraic operations or a monad, but treated specially by the language's top-level runtime, which invokes corresponding operating system functionality. While this approach works in practice, one wishes that the ingenuity of the language implementors were better supported by a more flexible methodology with a sound theoretical footing.
We address the issue by showing how to design a programming language based on runners of algebraic effects. We review runners, recast them as a programming construct, and present a calculus that captures the core ideas of programming with them. Through examples of runners we show how they capture both the interaction between the program and the external world, and encapsulation of programs in virtual environments that tightly control access to external resources and provide strong guarantees of proper resource finalization.
The entire community is invited to participate in a debate on the future of the conference system in theoretical computer science. This will provide a rare community-wide opportunity for us to discuss the strengths and weaknesses of our current system, and consider if we can do better. Questions will be asked by members of the audience, and answered by our panel members.
The scope of the debate is all aspects of our publishing and community traditions, characterised by prestige earned mostly through publication in competitive conferences, and frequent local and international travel. Possible topics for discussion include the need to publish in conferences for career progression, which usually involves burning carbon; wasted reviewing effort when good papers are rejected from highly competitive conferences; the extent of our responsibility as a community to respond to climate change; alternative publishing models, like the journal-focussed system used in mathematics; high costs of conference travel and registration; virtual conference advantages, disadvantages and best practice; improving equality, diversity and access; consequences and response to COVID-19; and the role of professional bodies. These topics have many tight relationships, and need to be discussed together to gain a full understanding of the issues involved.
Poll responses. Here are the results of the three anonymous polls that were held during the debate.
The following links provide further reading on topics related to the debate:
Joining the seminar. To join any OWLS seminar, visit the following link, up to 15 minutes before the posted start time:
You will be given the option to join through your web browser, or to launch the Zoom client if it is installed on your device (see below). For the best experience, we recommend using the client. When prompted for your name, please enter your real name, so that social interactions during the seminar and coffee breaks can be more natural.
Audio and video. We encourage all participants to enable their audio and video at all times (click "Use Device Audio" in the Zoom interface.) Don't worry about making noise and disrupting the proceedings accidentally; the Chairperson will ensure your audio is muted by default during the seminar. Having your audio and video enabled will allow other participants to see your face in the "Gallery" view, letting them know that you're taking part. It also gives you the option of asking a question, and of making best use of the "coffee break" sessions. For most users with good network access (such as a fast home broadband connection), there is no need to worry that having your audio and video enabled will degrade the experience; the technology platform ensures that the speaker's audio/video stream is prioritised at all times. However, those on slow connections may find it better to disable their audio and video.
Coffee breaks. Every OWLS seminar has two "coffee breaks", one starting 15 minutes before the posted start time of the seminar, and the second starting after the seminar is finished. To participate in these, feel free to join the meeting early, or to keep the meeting window open after the end of the talk. During these coffee break periods, participants will be automatically gathered into small groups, assigned at random; please introduce yourself to the other members of your group, and chat just like you would at a real conference. Remember to bring your own coffee! If you're not interested in chatting, please close the meeting window during the coffee break periods, so that people who want to chat are not grouped with those who do not.
During the seminar. If you'd like to ask a question, either during the seminar or in the question period at the end, click the "Participants" menu and select "Raise hand". The Chairperson may choose to interrupt the speaker and give your audio/video feed the focus, giving you the opportunity to ask your question verbally, or may instead decide to let the seminar continue. You may click "Lower hand" at any time to show you no longer wish to ask a question. To preserve the experience of a real face-to-face conference, there is no possibility of giving a written question, and the chat room is disabled. You also have the opportunity to give nonverbal feedback to the speaker by clicking the "speed up" or "slow down" buttons, also in the "Participants" menu.
Recordings. Most OWLS seminars will be recorded and uploaded to YouTube after the event. Only the audio/video of the chairperson, speaker, and questioners will be captured. If you prefer not to be recorded, do not ask a question. Of course, the organizers do not make any recordings of the coffee break sessions.
Inappropriate behaviour. The organizers make use of a range of security features to ensure that the seminar cannot be disrupted by participants who are not interested in constructive scientific interaction. During a seminar, if you observe ongoing inappropriate behaviour, please send a private message to the chairperson to point it out.
Seminars are hosted through the video conferencing system Zoom. Native clients are available to download for a wide range of platforms, including mobile devices:
For those who prefer not to use a client, a browser interface is also available.
If you have any feedback about OWLS, feel free to get in touch with an organizer.