-
Conservativity of Type Theory over Higher-order Arithmetic
Authors:
Benno van den Berg,
Daniël Otten
Abstract:
We investigate how much type theory is able to prove about the natural numbers. A classical result in this area shows that dependent type theory without any universes is conservative over Heyting Arithmetic (HA). We build on this result by showing that type theories with one level of universes are conservative over Higher-order Heyting Arithmetic (HAH). Although this clearly depends on the specifi…
▽ More
We investigate how much type theory is able to prove about the natural numbers. A classical result in this area shows that dependent type theory without any universes is conservative over Heyting Arithmetic (HA). We build on this result by showing that type theories with one level of universes are conservative over Higher-order Heyting Arithmetic (HAH). Although this clearly depends on the specific type theory, we show that the interpretation of logic also plays a major role. For proof-irrelevant interpretations, we will see that strong versions of type theory prove exactly the same higher-order arithmetical formulas as HAH. Conversely, proof-relevant interpretations prove strictly more second-order arithmetical formulas than HAH, however they still prove exactly the same first-order arithmetical formulas. Along the way, we investigate the different interpretations of logic in type theory, and to what extent dependent type theories can be seen as extensions of higher-order logic. We apply our results by proving De Jongh's theorem for type theory.
△ Less
Submitted 29 August, 2023;
originally announced August 2023.
-
Automatic Differentiation in Prolog
Authors:
Tom Schrijvers,
Birthe van den Berg,
Fabrizio Riguzzi
Abstract:
Automatic differentiation (AD) is a range of algorithms to compute the numeric value of a function's (partial) derivative, where the function is typically given as a computer program or abstract syntax tree. AD has become immensely popular as part of many learning algorithms, notably for neural networks. This paper uses Prolog to systematically derive gradient-based forward- and reverse-mode AD va…
▽ More
Automatic differentiation (AD) is a range of algorithms to compute the numeric value of a function's (partial) derivative, where the function is typically given as a computer program or abstract syntax tree. AD has become immensely popular as part of many learning algorithms, notably for neural networks. This paper uses Prolog to systematically derive gradient-based forward- and reverse-mode AD variants from a simple executable specification: evaluation of the symbolic derivative. Along the way we demonstrate that several Prolog features (DCGs, co-routines) contribute to the succinct formulation of the algorithm. We also discuss two applications in probabilistic programming that are enabled by our Prolog algorithms. The first is parameter learning for the Sum-Product Loop Language and the second consists of both parameter learning and variational inference for probabilistic logic programming.
△ Less
Submitted 13 May, 2023;
originally announced May 2023.
-
A Calculus for Scoped Effects & Handlers
Authors:
Roger Bosman,
Birthe van den Berg,
Wenhao Tang,
Tom Schrijvers
Abstract:
Algebraic effects & handlers have become a standard approach for side-effects in functional programming. Their modular composition with other effects and clean separation of syntax and semantics make them attractive to a wide audience. However, not all effects can be classified as algebraic; some need a more sophisticated handling. In particular, effects that have or create a delimited scope need…
▽ More
Algebraic effects & handlers have become a standard approach for side-effects in functional programming. Their modular composition with other effects and clean separation of syntax and semantics make them attractive to a wide audience. However, not all effects can be classified as algebraic; some need a more sophisticated handling. In particular, effects that have or create a delimited scope need special care, as their continuation consists of two parts-in and out of the scope-and their modular composition introduces additional complexity. These effects are called scoped and have gained attention by their growing applicability and adoption in popular libraries. While calculi have been designed with algebraic effects & handlers built in to facilitate their use, a calculus that supports scoped effects & handlers in a similar manner does not yet exist. This work fills this gap: we present $λ_{\mathit{sc}}$, a calculus with native support for both algebraic and scoped effects & handlers. It addresses the need for polymorphic handlers and explicit clauses for forwarding unknown scoped operations to other handlers. Our calculus is based on Eff, an existing calculus for algebraic effects, extended with Koka-style row polymorphism, and consists of a formal grammar, operational semantics, a (type-safe) type-and-effect system and type inference. We demonstrate $λ_{\mathit{sc}}$ on a range of examples.
△ Less
Submitted 5 March, 2024; v1 submitted 19 April, 2023;
originally announced April 2023.
-
A Framework for Higher-Order Effects & Handlers
Authors:
Birthe van den Berg,
Tom Schrijvers
Abstract:
Algebraic effects & handlers are a modular approach for modeling side-effects in functional programming. Their syntax is defined in terms of a signature of effectful operations, encoded as a functor, that are plugged into the free monad; their denotational semantics is defined by fold-style handlers that only interpret their part of the syntax and forward the rest. However, not all effects are alg…
▽ More
Algebraic effects & handlers are a modular approach for modeling side-effects in functional programming. Their syntax is defined in terms of a signature of effectful operations, encoded as a functor, that are plugged into the free monad; their denotational semantics is defined by fold-style handlers that only interpret their part of the syntax and forward the rest. However, not all effects are algebraic: some need to access an internal computation. For example, scoped effects distinguish between a computation in scope and out of scope; parallel effects parallellize over a computation, latent effects defer a computation. Separate definitions have been proposed for these higher-order effects and their corresponding handlers, often leading to expedient and complex monad definitions. In this work we propose a generic framework for higher-order effects, generalizing algebraic effects & handlers: a generic free monad with higher-order effect signatures and a corresponding interpreter. Specializing this higher-order syntax leads to various definitions of previously defined (scoped, parallel, latent) and novel (writer, bracketing) effects. Furthermore, we formally show our framework theoretically correct, also putting different effect instances on formal footing; a significant contribution for parallel, latent, writer and bracketing effects.
△ Less
Submitted 2 February, 2023;
originally announced February 2023.
-
Forward- or Reverse-Mode Automatic Differentiation: What's the Difference?
Authors:
Birthe van den Berg,
Tom Schrijvers,
James McKinna,
Alexander Vandenbroucke
Abstract:
Automatic differentiation (AD) has been a topic of interest for researchers in many disciplines, with increased popularity since its application to machine learning and neural networks. Although many researchers appreciate and know how to apply AD, it remains a challenge to truly understand the underlying processes. From an algebraic point of view, however, AD appears surprisingly natural: it orig…
▽ More
Automatic differentiation (AD) has been a topic of interest for researchers in many disciplines, with increased popularity since its application to machine learning and neural networks. Although many researchers appreciate and know how to apply AD, it remains a challenge to truly understand the underlying processes. From an algebraic point of view, however, AD appears surprisingly natural: it originates from the differentiation laws. In this work we use Algebra of Programming techniques to reason about different AD variants, leveraging Haskell to illustrate our observations. Our findings stem from three fundamental algebraic abstractions: (1) the notion of module over a semiring, (2) Nagata's construction of the 'idealization of a module', and (3) Kronecker's delta function, that together allow us to write a single-line abstract definition of AD. From this single-line definition, and by instantiating our algebraic structures in various ways, we derive different AD variants, that have the same extensional behaviour, but different intensional properties, mainly in terms of (asymptotic) computational complexity. We show the different variants equivalent by means of Kronecker isomorphisms, a further elaboration of our Haskell infrastructure which guarantees correctness by construction. With this framework in place, this paper seeks to make AD variants more comprehensible, taking an algebraic perspective on the matter.
△ Less
Submitted 9 August, 2023; v1 submitted 21 December, 2022;
originally announced December 2022.
-
Structured Handling of Scoped Effects: Extended Version
Authors:
Zhixuan Yang,
Marco Paviotti,
Nicolas Wu,
Birthe van den Berg,
Tom Schrijvers
Abstract:
Algebraic effects offer a versatile framework that covers a wide variety of effects. However, the family of operations that delimit scopes are not algebraic and are usually modelled as handlers, thus preventing them from being used freely in conjunction with algebraic operations. Although proposals for scoped operations exist, they are either ad-hoc and unprincipled, or too inconvenient for practi…
▽ More
Algebraic effects offer a versatile framework that covers a wide variety of effects. However, the family of operations that delimit scopes are not algebraic and are usually modelled as handlers, thus preventing them from being used freely in conjunction with algebraic operations. Although proposals for scoped operations exist, they are either ad-hoc and unprincipled, or too inconvenient for practical programming. This paper provides the best of both worlds: a theoretically-founded model of scoped effects that is convenient for implementation and reasoning. Our new model is based on an adjunction between a locally finitely presentable category and a category of functorial algebras. Using comparison functors between adjunctions, we show that our new model, an existing indexed model, and a third approach that simulates scoped operations in terms of algebraic ones have equal expressivity for handling scoped operations. We consider our new model to be the sweet spot between ease of implementation and structuredness. Additionally, our approach automatically induces fusion laws of handlers of scoped effects, which are useful for reasoning and optimisation.
△ Less
Submitted 25 January, 2022;
originally announced January 2022.
-
Latent Effects for Reusable Language Components: Extended Version
Authors:
Birthe van den Berg,
Tom Schrijvers,
Casper Bach-Poulsen,
Nicolas Wu
Abstract:
The development of programming languages can be quite complicated and costly. Hence, much effort has been devoted to the modular definition of language features that can be reused in various combinations to define new languages and experiment with their semantics. A notable outcome of these efforts is the algebra-based "datatypes a la carte" (DTC) approach. When combined with algebraic effects, DT…
▽ More
The development of programming languages can be quite complicated and costly. Hence, much effort has been devoted to the modular definition of language features that can be reused in various combinations to define new languages and experiment with their semantics. A notable outcome of these efforts is the algebra-based "datatypes a la carte" (DTC) approach. When combined with algebraic effects, DTC can model a wide range of common language features. Unfortunately, the current state of the art does not cover modular definitions of advanced control-flow mechanisms that defer execution to an appropriate point, such as call-by-name and call-by-need evaluation, as well as (multi-)staging. This paper defines latent effects, a generic class of such control-flow mechanisms. We demonstrate how function abstractions, lazy computations and a MetaML-like staging can all be expressed in a modular fashion using latent effects, and how they can be combined in various ways to obtain complex semantics. We provide a full Haskell implementation of our effects and handlers with a range of examples.
△ Less
Submitted 25 August, 2021;
originally announced August 2021.
-
Converse extensionality and apartness
Authors:
Benno van den Berg,
Robert Passmann
Abstract:
In this paper we try to find a computational interpretation for a strong form of extensionality, which we call "converse extensionality". Converse extensionality principles, which arise as the Dialectica interpretation of the axiom of extensionality, were first studied by Howard. In order to give a computational interpretation to these principles, we reconsider Brouwer's apartness relation, a stro…
▽ More
In this paper we try to find a computational interpretation for a strong form of extensionality, which we call "converse extensionality". Converse extensionality principles, which arise as the Dialectica interpretation of the axiom of extensionality, were first studied by Howard. In order to give a computational interpretation to these principles, we reconsider Brouwer's apartness relation, a strong constructive form of inequality. Formally, we provide a categorical construction to endow every typed combinatory algebra with an apartness relation. We then exploit that functions reflect apartness, in addition to preserving equality, to prove that the resulting categories of assemblies model a converse extensionality principle.
△ Less
Submitted 19 December, 2022; v1 submitted 26 March, 2021;
originally announced March 2021.
-
Quadratic type checking for objective type theory
Authors:
Benno van den Berg,
Martijn den Besten
Abstract:
We introduce a modification of standard Martin-Lof type theory in which we eliminate definitional equality and replace all computation rules by propositional equalities. We show that type checking for such a system can be done in quadratic time and that it has a natural homotopy-theoretic semantics.
We introduce a modification of standard Martin-Lof type theory in which we eliminate definitional equality and replace all computation rules by propositional equalities. We show that type checking for such a system can be done in quadratic time and that it has a natural homotopy-theoretic semantics.
△ Less
Submitted 1 February, 2021;
originally announced February 2021.
-
Effective Kan fibrations in simplicial sets
Authors:
Benno van den Berg,
Eric Faber
Abstract:
We introduce the notion of an effective Kan fibration, a new mathematical structure that can be used to study simplicial homotopy theory. Our main motivation is to make simplicial homotopy theory suitable for homotopy type theory. Effective Kan fibrations are maps of simplicial sets equipped with a structured collection of chosen lifts that satisfy certain non-trivial properties. This contrasts wi…
▽ More
We introduce the notion of an effective Kan fibration, a new mathematical structure that can be used to study simplicial homotopy theory. Our main motivation is to make simplicial homotopy theory suitable for homotopy type theory. Effective Kan fibrations are maps of simplicial sets equipped with a structured collection of chosen lifts that satisfy certain non-trivial properties. This contrasts with the ordinary, unstructured notion of a Kan fibration. We show that fundamental properties of Kan fibrations can be extended to explicit constructions on effective Kan fibrations. In particular, we give a constructive (explicit) proof showing that effective Kan fibrations are stable under push forward, or fibred exponentials. This is known to be impossible for ordinary Kan fibrations. We further show that effective Kan fibrations are local, or completely determined by their fibres above representables. We also give an (ineffective) proof saying that the maps which can be equipped with the structure of an effective Kan fibration are precisely the ordinary Kan fibrations. Hence implicitly, both notions still describe the same homotopy theory. By showing that the effective Kan fibrations combine all these properties, we solve an open problem in homotopy type theory. In this way our work provides a first step in giving a constructive account of Voevodsky's model of univalent type theory in simplicial sets.
△ Less
Submitted 30 April, 2022; v1 submitted 26 September, 2020;
originally announced September 2020.
-
Analyzing Partitioned FAIR Health Data Responsibly
Authors:
Chang Sun,
Lianne Ippel,
Birgit Wouters,
Johan van Soest,
Alexander Malic,
Onaopepo Adekunle,
Bob van den Berg,
Marco Puts,
Ole Mussmann,
Annemarie Koster,
Carla van der Kallen,
David Townend,
Andre Dekker,
Michel Dumontier
Abstract:
It is widely anticipated that the use of health-related big data will enable further understanding and improvements in human health and wellbeing. Our current project, funded through the Dutch National Research Agenda, aims to explore the relationship between the development of diabetes and socio-economic factors such as lifestyle and health care utilization. The analysis involves combining data f…
▽ More
It is widely anticipated that the use of health-related big data will enable further understanding and improvements in human health and wellbeing. Our current project, funded through the Dutch National Research Agenda, aims to explore the relationship between the development of diabetes and socio-economic factors such as lifestyle and health care utilization. The analysis involves combining data from the Maastricht Study (DMS), a prospective clinical study, and data collected by Statistics Netherlands (CBS) as part of its routine operations. However, a wide array of social, legal, technical, and scientific issues hinder the analysis. In this paper, we describe these challenges and our progress towards addressing them.
△ Less
Submitted 2 December, 2018;
originally announced December 2018.
-
Univalent polymorphism
Authors:
Benno van den Berg
Abstract:
We show that Martin Hyland's effective topos can be exhibited as the homotopy category of a path category $\mathbb{EFF}$. Path categories are categories of fibrant objects in the sense of Brown satisfying two additional properties and as such provide a context in which one can interpret many notions from homotopy theory and Homotopy Type Theory. Within the path category $\mathbb{EFF}$ one can iden…
▽ More
We show that Martin Hyland's effective topos can be exhibited as the homotopy category of a path category $\mathbb{EFF}$. Path categories are categories of fibrant objects in the sense of Brown satisfying two additional properties and as such provide a context in which one can interpret many notions from homotopy theory and Homotopy Type Theory. Within the path category $\mathbb{EFF}$ one can identify a class of discrete fibrations which is closed under push forward along arbitrary fibrations (in other words, this class is polymorphic or closed under impredicative quantification) and satisfies propositional resizing. This class does not have a univalent representation, but if one restricts to those discrete fibrations whose fibres are propositions in the sense of Homotopy Type Theory, then it does. This means that, modulo the usual coherence problems, it can be seen as a model of the Calculus of Constructions with a univalent type of propositions. We will also build a more complicated path category in which the class of discrete fibrations whose fibres are sets in the sense of Homotopy Type Theory has a univalent representation, which means that this will be a model of the Calculus of Constructions with a univalent type of sets.
△ Less
Submitted 1 August, 2018; v1 submitted 27 March, 2018;
originally announced March 2018.
-
A homotopy-theoretic model of function extensionality in the effective topos
Authors:
Daniil Frumin,
Benno van den Berg
Abstract:
We present a way of constructing a Quillen model structure on a full subcategory of an elementary topos, starting with an interval object with connections and a certain dominance. The advantage of this method is that it does not require the underlying topos to be cocomplete. The resulting model category structure gives rise to a model of homotopy type theory with identity types, $Σ$- and $Π$-types…
▽ More
We present a way of constructing a Quillen model structure on a full subcategory of an elementary topos, starting with an interval object with connections and a certain dominance. The advantage of this method is that it does not require the underlying topos to be cocomplete. The resulting model category structure gives rise to a model of homotopy type theory with identity types, $Σ$- and $Π$-types, and functional extensionality.
We apply the method to the effective topos with the interval object $\nabla 2$. In the resulting model structure we identify uniform inhabited objects as contractible objects, and show that discrete objects are fibrant. Moreover, we show that the unit of the discrete reflection is a homotopy equivalence and the homotopy category of fibrant assemblies is equivalent to the category of modest sets. We compare our work with the path object category construction on the effective topos by Jaap van Oosten.
△ Less
Submitted 12 March, 2018; v1 submitted 29 January, 2017;
originally announced January 2017.
