The Way of the Dagger

Total Page:16

File Type:pdf, Size:1020Kb

The Way of the Dagger The Way of the Dagger Martti Karvonen I V N E R U S E I T H Y T O H F G E R D I N B U Doctor of Philosophy Laboratory for Foundations of Computer Science School of Informatics University of Edinburgh 2018 (Graduation date: July 2019) Abstract A dagger category is a category equipped with a functorial way of reversing morph- isms, i.e. a contravariant involutive identity-on-objects endofunctor. Dagger categor- ies with additional structure have been studied under different names in categorical quantum mechanics, algebraic field theory and homological algebra, amongst others. In this thesis we study the dagger in its own right and show how basic category theory adapts to dagger categories. We develop a notion of a dagger limit that we show is suitable in the following ways: it subsumes special cases known from the literature; dagger limits are unique up to unitary isomorphism; a wide class of dagger limits can be built from a small selection of them; dagger limits of a fixed shape can be phrased as dagger adjoints to a diagonal functor; dagger limits can be built from ordinary limits in the presence of polar decomposition; dagger limits commute with dagger colimits in many cases. Using cofree dagger categories, the theory of dagger limits can be leveraged to provide an enrichment-free understanding of limit-colimit coincidences in ordinary category theory. We formalize the concept of an ambilimit, and show that it captures known cases. As a special case, we show how to define biproducts up to isomorphism in an arbitrary category without assuming any enrichment. Moreover, the limit-colimit coincidence from domain theory can be generalized to the unenriched setting and we show that, under suitable assumptions, a wide class of endofunctors has canonical fixed points. The theory of monads on dagger categories works best when all structure respects the dagger: the monad and adjunctions should preserve the dagger, and the monad and its algebras should satisfy the so-called Frobenius law. Then any monad resolves as an adjunction, with extremal solutions given by the categories of Kleisli and Frobenius- Eilenberg-Moore algebras, which again have a dagger. We use dagger categories to study reversible computing. Specifically, we model re- versible effects by adapting Hughes’ arrows to dagger arrows and inverse arrows. This captures several fundamental reversible effects, including serialization and mutable store computations. Whereas arrows are monoids in the category of profunctors, dag- ger arrows are involutive monoids in the category of profunctors, and inverse arrows satisfy certain additional properties. These semantics inform the design of functional reversible programs supporting side-effects. iii Lay Summary This thesis is primarily about the mathematics of so-called dagger categories. For now, think of a category as a collection of systems A;B;C;::: and processes between them. Processes can be done after each other, and for any system A there should be a trivial process idA that amounts to doing nothing. In a dagger category there is an additional piece of structure – any process f : A ! B has a counterpart process f † : B ! A going in the opposite direction. One might think of f † as some kind of a reversal of the process f , and this motivates the rules this reversal is supposed to obey, namely • reversing the trivial process of any system A results in the trivial process, i.e. † idA = idA • reversing a process twice results in the original process, i.e. f †† = f • if one reverses the process “g after f ”, one might as well have reversed each process individually and done them after each other in the opposite order. More succinctly, (g ◦ f )† = f † ◦ g† Any mathematical structure obeying these rules is a dagger category. For a concrete example, let us consider the alphabet. Letters can be written after one another, resulting in strings such as “asdf” and “cat”. Strings can be composed by writing them one after the other, so “world” composed after “hello” becomes ‘helloworld”. The empty string is a neutral element with respect to composition, i.e. appending or prepending it to a string doesn’t change the string in question. What makes this into a dagger category is the fact that strings can be reversed. Reversing the empty string results in an empty string. Reversing a string twice results in the original string. Reversing a composite string is the same thing as composing the reversals in the opposite order. Hence the rules above are indeed obeyed by strings of letters. Dagger categories arise independently both in physics and computation, and also at their intersection in quantum computing. However, so far a systematic study of dagger categories has been missing, and this thesis fills the gap. There are various mathem- atical questions and notions people study in the context of ordinary categories, such as (co)limits, which consider well-behaved ways of building new systems from old ones, or monads and arrows, which can be used to enlarge the collection of processes available. This thesis shows how to understand such topics in the context of dagger categories. iv Acknowledgements First and foremost I need to thank my first supervisor Chris Heunen for his guid- ance throughout the years. Besides helping in choosing a topic, nudging me in the right direction when I got stuck, giving advice on literature to read and advising with numerous smaller issues that come about in academic life, he has helped me to grow into a researcher who can hopefully make all these decisions alone. Without him you would be reading an empty thesis1. I must also thank my second supervisor Tom Leinster. His intuition, whether per- taining to topics described in this thesis or to mathematics in general, has not only been helpful but also a delight to observe. I hope that I have been able to absorb even a fraction of his way of thinking. I am very thankful for my examiners Peter Selinger and David Jordan. The time they invested into reading this thesis with care resulted in insightful comments and feedback that has greatly improved this thesis. I wish to thank Gordon Plotkin for very helpful comments during my yearly re- views and on an earlier draft. It is truly a pleasure to see how a computer scientist thinks about category theory. Similarly, I must thank Robin Kaarsgaard not only for his helpful comments but also for the work we have done together – hopefully this was only a start. To the extent that the thesis falls under theoretical computer science, it is because of Gordon and Robin. I am indebted for Jamie Vicary both for coining the name “way of the dagger” to describe the general philosophy when working with dagger categories, and for letting me use it as a title. More widely, I am thankful for the whole community centred around the conference Quantum Physics and Logic – it is a great research community to be a part of, and a great source of friends and collaborators. Before moving to friends and family, I wish to thank various organizations that have helped me. I wish to thank the Osk. Huttunen foundation for financially sup- porting me during my PhD. The money that Vais¨ al¨ a¨ foundation kindly let me use as a travel grant was very useful for letting me become a part of an international scientific community and I learned a lot on the way. COST Action IC1405 made possible a trip to Copenhagen, without which this thesis would be shorter. My host institute LFCS was not only a great academic community, but also helped me with a funding gap at the start and some travel money during the later years. 1A variant of [144] v Life in Edinburgh would have been much duller without Toms, Kima and their friends. Climbing with Orfeas has provided me with a much needed counterbalance to research. My summer breaks in Finland were mostly timed so that I could visit KKT, so I have to thank everyone involved for that particular getaway and the support therein. I must also thank Tuomo, Sakari and IHRA more generally for their sustained friendship. I am grateful for Pentti, Pirjo and Pekka, not only for the usual things loving fam- ilies provide, but also for being crucial for my intellectual development during my formative years and supportive once I got interested in mathematics. All of my sib- lings, half or otherwise, have been crucial to me growing up to be the person I am. There is no way words can express the gratitude I have for Esma, so I will not even try. vi Declaration I declare that this thesis was composed by myself, that the work contained herein is my own except where explicitly stated otherwise in the text, and that this work has not been submitted for any other degree or professional qualification except as specified. (Martti Karvonen) vii Contents 1 Introduction 1 1.1 Why dagger categories . .1 1.2 Historical Background . .2 1.3 Outline, results and content available elsewhere . .5 1.4 Prerequisites . .6 1.5 On notation . .7 2 Background 9 2.1 Dagger categories . .9 2.2 Graphical calculus . 16 3 Equivalences and adjunctions 21 3.1 Dagger equivalences . 21 3.2 Dagger adjunctions . 25 4 Limits 29 4.1 Introduction . 29 4.2 Dagger limits . 30 4.3 Uniqueness up to unitary isomorphism . 37 4.4 Completeness . 43 4.5 Global dagger limits . 50 4.6 Dagger adjoint functors . 53 4.7 Polar decomposition . 55 4.8 Commutativity of limits and colimits . 64 5 Ordinary limit-colimit coincidences via dagger limits 73 5.1 Ambilimits .
Recommended publications
  • Experience Implementing a Performant Category-Theory Library in Coq
    Experience Implementing a Performant Category-Theory Library in Coq Jason Gross, Adam Chlipala, and David I. Spivak Massachusetts Institute of Technology, Cambridge, MA, USA [email protected], [email protected], [email protected] Abstract. We describe our experience implementing a broad category- theory library in Coq. Category theory and computational performance are not usually mentioned in the same breath, but we have needed sub- stantial engineering effort to teach Coq to cope with large categorical constructions without slowing proof script processing unacceptably. In this paper, we share the lessons we have learned about how to repre- sent very abstract mathematical objects and arguments in Coq and how future proof assistants might be designed to better support such rea- soning. One particular encoding trick to which we draw attention al- lows category-theoretic arguments involving duality to be internalized in Coq's logic with definitional equality. Ours may be the largest Coq development to date that uses the relatively new Coq version developed by homotopy type theorists, and we reflect on which new features were especially helpful. Keywords: Coq · category theory · homotopy type theory · duality · performance 1 Introduction Category theory [36] is a popular all-encompassing mathematical formalism that casts familiar mathematical ideas from many domains in terms of a few unifying concepts. A category can be described as a directed graph plus algebraic laws stating equivalences between paths through the graph. Because of this spar- tan philosophical grounding, category theory is sometimes referred to in good humor as \formal abstract nonsense." Certainly the popular perception of cat- egory theory is quite far from pragmatic issues of implementation.
    [Show full text]
  • A Subtle Introduction to Category Theory
    A Subtle Introduction to Category Theory W. J. Zeng Department of Computer Science, Oxford University \Static concepts proved to be very effective intellectual tranquilizers." L. L. Whyte \Our study has revealed Mathematics as an array of forms, codify- ing ideas extracted from human activates and scientific problems and deployed in a network of formal rules, formal definitions, formal axiom systems, explicit theorems with their careful proof and the manifold in- terconnections of these forms...[This view] might be called formal func- tionalism." Saunders Mac Lane Let's dive right in then, shall we? What can we say about the array of forms MacLane speaks of? Claim (Tentative). Category theory is about the formal aspects of this array of forms. Commentary: It might be tempting to put the brakes on right away and first grab hold of what we mean by formal aspects. Instead, rather than trying to present \the formal" as the object of study and category theory as our instrument, we would nudge the reader to consider another perspective. Category theory itself defines formal aspects in much the same way that physics defines physical concepts or laws define legal (and correspondingly illegal) aspects: it embodies them. When we speak of what is formal/physical/legal we inescapably speak of category/physical/legal theory, and vice versa. Thus we pass to the somewhat grammatically awkward revision of our initial claim: Claim (Tentative). Category theory is the formal aspects of this array of forms. Commentary: Let's unpack this a little bit. While individual forms are themselves tautologically formal, arrays of forms and everything else networked in a system lose this tautological formality.
    [Show full text]
  • Compactly Accessible Categories and Quantum Key Distribution
    COMPACTLY ACCESSIBLE CATEGORIES AND QUANTUM KEY DISTRIBUTION CHRIS HEUNEN Institute for Computing and Information Sciences, Radboud University, Nijmegen, the Netherlands Abstract. Compact categories have lately seen renewed interest via applications to quan- tum physics. Being essentially finite-dimensional, they cannot accomodate (co)limit-based constructions. For example, they cannot capture protocols such as quantum key distribu- tion, that rely on the law of large numbers. To overcome this limitation, we introduce the notion of a compactly accessible category, relying on the extra structure of a factorisation system. This notion allows for infinite dimension while retaining key properties of compact categories: the main technical result is that the choice-of-duals functor on the compact part extends canonically to the whole compactly accessible category. As an example, we model a quantum key distribution protocol and prove its correctness categorically. 1. Introduction Compact categories were first introduced in 1972 as a class of examples in the con- text of the coherence problem [Kel72]. They were subsequently studied first categori- cally [Day77, KL80], and later in relation to linear logic [See89]. Interest has rejuvenated since the exhibition of another aspect: compact categories provide a semantics for quantum computation [AC04, Sel07]. The main virtue of compact categories as models of quantum computation is that from very few axioms, surprisingly many consequences ensue that were postulates explicitly in the traditional Hilbert space formalism, e.g. scalars [Abr05]. More- over, the connection to linear logic provides quantum computation with a resource sensitive type theory of its own [Dun06]. Much of the structure of compact categories is due to a seemingly ingrained ‘finite- dimensionality'.
    [Show full text]
  • The Morita Theory of Quantum Graph Isomorphisms
    Commun. Math. Phys. Communications in Digital Object Identifier (DOI) https://doi.org/10.1007/s00220-018-3225-6 Mathematical Physics The Morita Theory of Quantum Graph Isomorphisms Benjamin Musto, David Reutter, Dominic Verdon Department of Computer Science, University of Oxford, Oxford, UK. E-mail: [email protected]; [email protected]; [email protected] Received: 31 January 2018 / Accepted: 11 June 2018 © The Author(s) 2018 Abstract: We classify instances of quantum pseudo-telepathy in the graph isomorphism game, exploiting the recently discovered connection between quantum information and the theory of quantum automorphism groups. Specifically, we show that graphs quantum isomorphic to a given graph are in bijective correspondence with Morita equivalence classes of certain Frobenius algebras in the category of finite-dimensional representations of the quantum automorphism algebra of that graph. We show that such a Frobenius algebra may be constructed from a central type subgroup of the classical automorphism group, whose action on the graph has coisotropic vertex stabilisers. In particular, if the original graph has no quantum symmetries, quantum isomorphic graphs are classified by such subgroups. We show that all quantum isomorphic graph pairs corresponding to a well-known family of binary constraint systems arise from this group-theoretical construction. We use our classification to show that, of the small order vertex-transitive graphs with no quantum symmetry, none is quantum isomorphic to a non-isomorphic graph. We show that this is in fact asymptotically almost surely true of all graphs. Contents 1. Introduction ................................. 2. Background ................................. 2.1 String diagrams, Frobenius monoids and Gelfand duality ......
    [Show full text]
  • Categories of Quantum and Classical Channels (Extended Abstract)
    Categories of Quantum and Classical Channels (extended abstract) Bob Coecke∗ Chris Heunen† Aleks Kissinger∗ University of Oxford, Department of Computer Science fcoecke,heunen,[email protected] We introduce the CP*–construction on a dagger compact closed category as a generalisation of Selinger’s CPM–construction. While the latter takes a dagger compact closed category and forms its category of “abstract matrix algebras” and completely positive maps, the CP*–construction forms its category of “abstract C*-algebras” and completely positive maps. This analogy is justified by the case of finite-dimensional Hilbert spaces, where the CP*–construction yields the category of finite-dimensional C*-algebras and completely positive maps. The CP*–construction fully embeds Selinger’s CPM–construction in such a way that the objects in the image of the embedding can be thought of as “purely quantum” state spaces. It also embeds the category of classical stochastic maps, whose image consists of “purely classical” state spaces. By allowing classical and quantum data to coexist, this provides elegant abstract notions of preparation, measurement, and more general quantum channels. 1 Introduction One of the motivations driving categorical treatments of quantum mechanics is to place classical and quantum systems on an equal footing in a single category, so that one can study their interactions. The main idea of categorical quantum mechanics [1] is to fix a category (usually dagger compact) whose ob- jects are thought of as state spaces and whose morphisms are evolutions. There are two main variations. • “Dirac style”: Objects form pure state spaces, and isometric morphisms form pure state evolutions.
    [Show full text]
  • Compact Inverse Categories 11
    COMPACT INVERSE CATEGORIES ROBIN COCKETT AND CHRIS HEUNEN Abstract. The Ehresmann-Schein-Nambooripad theorem gives a structure theorem for inverse monoids: they are inductive groupoids. A particularly nice case due to Jarek is that commutative inverse monoids become semilattices of abelian groups. It has also been categorified by DeWolf-Pronk to a structure theorem for inverse categories as locally complete inductive groupoids. We show that in the case of compact inverse categories, this takes the particularly nice form of a semilattice of compact groupoids. Moreover, one-object com- pact inverse categories are exactly commutative inverse monoids. Compact groupoids, in turn, are determined in particularly simple terms of 3-cocycles by Baez-Lauda. 1. Introduction Inverse monoids model partial symmetry [24], and arise naturally in many com- binatorial constructions [8]. The easiest example of an inverse monoid is perhaps a group. There is a structure theorem for inverse monoids, due to Ehresmann-Schein- Nambooripad [9, 10, 27, 26], that exhibits them as inductive groupoids. The latter are groupoids internal to the category of partially ordered sets with certain ex- tra requirements. By a result of Jarek [19], the inductive groupoids corresponding to commutative inverse monoids can equivalently be described as semilattices of abelian groups. A natural typed version of an inverse monoid is an inverse category [22, 6]. This notion can for example model partial reversible functional programs [12]. The eas- iest example of an inverse category is perhaps a groupoid. DeWolf-Pronk have generalised the ESN theorem to inverse categories, exhibiting them as locally com- plete inductive groupoids. This paper investigates `the commutative case', thus fitting in the bottom right cell of Figure 1.
    [Show full text]
  • Rewriting Structured Cospans: a Syntax for Open Systems
    UNIVERSITY OF CALIFORNIA RIVERSIDE Rewriting Structured Cospans: A Syntax For Open Systems A Dissertation submitted in partial satisfaction of the requirements for the degree of Doctor of Philosophy in Mathematics by Daniel Cicala June 2019 Dissertation Committee: Dr. John C. Baez, Chairperson Dr. Wee Liang Gan Dr. Jacob Greenstein Copyright by Daniel Cicala 2019 The Dissertation of Daniel Cicala is approved: Committee Chairperson University of California, Riverside Acknowledgments First and foremost, I would like to thank my advisor John Baez. In these past few years, I have learned more than I could have imagined about mathematics and the job of doing mathematics. I also want to thank the past and current Baez Crew for the many wonderful discussions. I am indebted to Math Department at the University of California, Riverside, which has afforded me numerous opportunities to travel to conferences near and far. Almost certainly, I would never have had a chance to pursue my doctorate had it not been for my parents who were there for me through every twist and turn on this, perhaps, too scenic route that I traveled. Most importantly, this project would have been impossible without the full-hearted support of my love, Elizabeth. I would also like to acknowledge the previously published material in this disser- tation. The interchange law in Section 3.1 was published in [15]. The material in Sections 3.2 and 3.3 appear in [16]. Also, the ZX-calculus example in Section 4.3 appears in [18]. iv Elizabeth. It’s finally over, baby! v ABSTRACT OF THE DISSERTATION Rewriting Structured Cospans: A Syntax For Open Systems by Daniel Cicala Doctor of Philosophy, Graduate Program in Mathematics University of California, Riverside, June 2019 Dr.
    [Show full text]
  • Most Human Things Go in Pairs. Alcmaeon, ∼ 450 BC
    Most human things go in pairs. Alcmaeon, 450 BC ∼ true false good bad right left up down front back future past light dark hot cold matter antimatter boson fermion How can we formalize a general concept of duality? The Chinese tried yin-yang theory, which inspired Leibniz to develop binary notation, which in turn underlies digital computation! But what's the state of the art now? In category theory the fundamental duality is the act of reversing an arrow: • ! • • • We use this to model switching past and future, false and true, small and big... Every category has an opposite op, where the arrows are C C reversed. This is a symmetry of the category of categories: op : Cat Cat ! and indeed the only nontrivial one: Aut(Cat) = Z=2 In logic, the simplest duality is negation. It's order-reversing: if P implies Q then Q implies P : : and|ignoring intuitionism!|it's an involution: P = P :: Thus if is a category of propositions and proofs, we C expect a functor: : op : C ! C with a natural isomorphism: 2 = 1 : ∼ C This has two analogues in quantum theory. One shows up already in the category of finite-dimensional vector spaces, FinVect. Every vector space V has a dual V ∗. Taking the dual is contravariant: if f : V W then f : W V ! ∗ ∗ ! ∗ and|ignoring infinite-dimensional spaces!|it's an involution: V ∗∗ ∼= V This kind of duality is captured by the idea of a -autonomous ∗ category. Recall that a symmetric monoidal category is roughly a category with a unit object I and a tensor product C 2 C : ⊗ C × C ! C that is unital, associative and commutative up to coherent natural isomorphisms.
    [Show full text]
  • Experience Implementing a Performant Category-Theory Library in Coq
    Experience Implementing a Performant Category-Theory Library in Coq Jason Gross, Adam Chlipala, and David I. Spivak Massachusetts Institute of Technology, Cambridge, MA, USA [email protected], [email protected], [email protected] Abstract. We describe our experience implementing a broad category- theory library in Coq. Category theory and computational performance are not usually mentioned in the same breath, but we have needed sub- stantial engineering effort to teach Coq to cope with large categorical constructions without slowing proof script processing unacceptably. In this paper, we share the lessons we have learned about how to repre- sent very abstract mathematical objects and arguments in Coq and how future proof assistants might be designed to better support such rea- soning. One particular encoding trick to which we draw attention al- lows category-theoretic arguments involving duality to be internalized in Coq's logic with definitional equality. Ours may be the largest Coq development to date that uses the relatively new Coq version developed by homotopy type theorists, and we reflect on which new features were especially helpful. Keywords: Coq · category theory · homotopy type theory · duality · performance 1 Introduction Category theory [14] is a popular all-encompassing mathematical formalism that casts familiar mathematical ideas from many domains in terms of a few unifying concepts. A category can be described as a directed graph plus algebraic laws stating equivalences between paths through the graph. Because of this spar- tan philosophical grounding, category theory is sometimes referred to in good humor as \formal abstract nonsense." Certainly the popular perception of cat- egory theory is quite far from pragmatic issues of implementation.
    [Show full text]
  • Categories of Relations As Models of Quantum Theory
    Categories of relations as models of quantum theory Chris Heunen∗ and Sean Tull† fheunen,[email protected] University of Oxford, Department of Computer Science Categories of relations over a regular category form a family of models of quantum theory. Using regular logic, many properties of relations over sets lift to these models, including the correspon- dence between Frobenius structures and internal groupoids. Over compact Hausdorff spaces, this lifting gives continuous symmetric encryption. Over a regular Mal’cev category, this correspondence gives a characterization of categories of completely positive maps, enabling the formulation of quan- tum features. These models are closer to Hilbert spaces than relations over sets in several respects: Heisenberg uncertainty, impossibility of broadcasting, and behavedness of rank one morphisms. 1 Introduction Many features of quantum theory can be abstracted to arbitrary compact dagger categories [1]. Thus we can compare models of quantum theory in a unified setting, and look for features that distinguish the category FHilb of finite-dimensional Hilbert spaces that forms the traditional model. The category Rel of sets and relations is an alternative model. It exhibits many features considered typical of quantum theory [2], but also refutes presumed correspondences between them [15, 21]. However, apart from Rel and its subcategory corresponding to Spekkens’ toy model [14], few alternative models have been studied in detail. We consider a new family of models by generalizing to categories Rel(C) of relations over an arbi- trary regular category C, including any algebraic category like that of groups, any abelian category like that of vector spaces, and any topos like that of sets.
    [Show full text]
  • A Survey of Graphical Languages for Monoidal Categories
    5 Traced categories 32 5.1 Righttracedcategories . 33 5.2 Planartracedcategories. 34 5.3 Sphericaltracedcategories . 35 A survey of graphical languages for monoidal 5.4 Spacialtracedcategories . 36 categories 5.5 Braidedtracedcategories . 36 5.6 Balancedtracedcategories . 39 5.7 Symmetrictracedcategories . 40 Peter Selinger 6 Products, coproducts, and biproducts 41 Dalhousie University 6.1 Products................................. 41 6.2 Coproducts ............................... 43 6.3 Biproducts................................ 44 Abstract 6.4 Traced product, coproduct, and biproduct categories . ........ 45 This article is intended as a reference guide to various notions of monoidal cat- 6.5 Uniformityandregulartrees . 47 egories and their associated string diagrams. It is hoped that this will be useful not 6.6 Cartesiancenter............................. 48 just to mathematicians, but also to physicists, computer scientists, and others who use diagrammatic reasoning. We have opted for a somewhat informal treatment of 7 Dagger categories 48 topological notions, and have omitted most proofs. Nevertheless, the exposition 7.1 Daggermonoidalcategories . 50 is sufficiently detailed to make it clear what is presently known, and to serve as 7.2 Otherprogressivedaggermonoidalnotions . ... 50 a starting place for more in-depth study. Where possible, we provide pointers to 7.3 Daggerpivotalcategories. 51 more rigorous treatments in the literature. Where we include results that have only 7.4 Otherdaggerpivotalnotions . 54 been proved in special cases, we indicate this in the form of caveats. 7.5 Daggertracedcategories . 55 7.6 Daggerbiproducts............................ 56 Contents 8 Bicategories 57 1 Introduction 2 9 Beyond a single tensor product 58 2 Categories 5 10 Summary 59 3 Monoidal categories 8 3.1 (Planar)monoidalcategories . 9 1 Introduction 3.2 Spacialmonoidalcategories .
    [Show full text]
  • Categories with Fuzzy Sets and Relations
    Categories with Fuzzy Sets and Relations John Harding, Carol Walker, Elbert Walker Department of Mathematical Sciences New Mexico State University Las Cruces, NM 88003, USA jhardingfhardy,[email protected] Abstract We define a 2-category whose objects are fuzzy sets and whose maps are relations subject to certain natural conditions. We enrich this category with additional monoidal and involutive structure coming from t-norms and negations on the unit interval. We develop the basic properties of this category and consider its relation to other familiar categories. A discussion is made of extending these results to the setting of type-2 fuzzy sets. 1 Introduction A fuzzy set is a map A : X ! I from a set X to the unit interval I. Several authors [2, 6, 7, 20, 22] have considered fuzzy sets as a category, which we will call FSet, where a morphism from A : X ! I to B : Y ! I is a function f : X ! Y that satisfies A(x) ≤ (B◦f)(x) for each x 2 X. Here we continue this path of investigation. The underlying idea is to lift additional structure from the unit interval, such as t-norms, t-conorms, and negations, to provide additional structure on the category. Our eventual aim is to provide a setting where processes used in fuzzy control can be abstractly studied, much in the spirit of recent categorical approaches to processes used in quantum computation [1]. Order preserving structure on I, such as t-norms and conorms, lifts to provide additional covariant structure on FSet. In fact, each t-norm T lifts to provide a symmetric monoidal tensor ⊗T on FSet.
    [Show full text]