Searcharxiv⌕ Search

arXiv subjects

David I. Spivak

Publications and source records attributed to David I. Spivak.

At least 37 records · Page 2Linked to original sources
arXiv2023

Dynamic Operads, Dynamic Categories: From Deep Learning to Prediction Markets

Natural organized systems adapt to internal and external pressures and this happens at all levels of the abstraction hierarchy. Wanting to think clearly about this idea motivates our paper, and so the idea is elaborated extensively in the introduction, which should be broadly accessible to a philosophically-interested audience. In the remaining sections, we turn to more compressed category theory. We define the monoidal double category Org of dynamic organizations, we provide definitions of Org-enriched, or dynamic, categorical structures -- e.g. dynamic categories, operads, and monoidal categories -- and we show how they instantiate the motivating philosophical ideas. We give two examples of dynamic categorical structures: prediction markets as a dynamic operad and deep learning as a dynamic monoidal category.

math.CT↗
arXiv2023

Polynomial Functors and Shannon Entropy

Past work shows that one can associate a notion of Shannon entropy to a Dirichlet polynomial, regarded as an empirical distribution. Indeed, entropy can be extracted from any d:Dir by a two-step process, where the first step is a rig homomorphism out of Dir, the *set* of Dirichlet polynomials, with rig structure given by standard addition and multiplication. In this short note, we show that this rig homomorphism can be upgraded to a rig *functor*, when we replace the set of Dirichlet polynomials by the *category* of ordinary (Cartesian) polynomials. In the Cartesian case, the process has three steps. The first step is a rig functor PolyCart -> Poly sending a polynomial p to (dp)y, where dp is the derivative of p. The second is a rig functor Poly -> Set x Set^op, sending a polynomial q to the pair (q(1),Gamma(q)), where Gamma(q)=Poly(q,y) can be interpreted as the global sections of q viewed as a bundle, and q(1) as its base. To make this precise we define what appears to be a new distributive monoidal structure on Set x Set^op, which can be understood geometrically in terms of rectangles. The last step, as for Dirichlet polynomials, is simply to extract the entropy as a real number from a pair of sets (A,B); it is given by log A - log B^(1/A) and can be thought of as the log aspect ratio of the rectangle.

math.CT↗
arXiv2023

Structures on Categories of Polynomials

We define the monoidal category $(Poly_E,y,\triangleleft)$ of polynomials under composition in any category $E$ with finite limits, including both cartesian and vertical morphisms of polynomials, and generalize to this setting the Dirichlet tensor product of polynomials $\otimes$, duoidality of $\otimes$ and $\triangleleft$, closure of $\otimes$, and coclosures of $\triangleleft$. We also prove that $\triangleleft$-comonoids in $Poly_E$ are precisely the internal categories in $E$ whose source morphism is exponentiable, generalizing a result of Ahman-Uustalu equating categories with polynomial comonads, and show that coalgebras in this setting correspond to internal copresheaves. Finally, the double category of ``typed'' polynomials in $E$ is recovered using $\triangleleft$-bicomodules in $Poly_E$.

math.CT↗
arXiv2023

Dynamic Tracing: a graphical language for rewriting protocols

The category Set_* of sets and partial functions is well-known to be traced monoidal, meaning that a partial function S+U -/-> T+U can be coherently transformed into a partial function S -/-> T. This transformation is generally described in terms of an implicit procedure that must be run. We make this procedure explicit by enriching the traced category in Cat#, the symmetric monoidal category of categories and cofunctors: each hom-category has such procedures as objects, and advancement through the procedures as arrows. We also generalize to traced Kleisli categories beyond Set_*, providing a conjectural trace operator for the Kleisli category of any polynomial monad of the form t+1. The main motivation for this work is to give a formal and graphical syntax for performing sophisticated computations powered by graph rewriting, which is itself a graphical language for data transformation.

cs.LO↗
arXiv2022

Temporal Landscapes: A Graphical Logic of Behavior

We present an elementary introduction to a new logic for reasoning about behaviors that occur over time. This logic is based on temporal type theory. The syntax of the logic is similar to the usual first-order logic; what differs is the notion of truth value. Instead of reasoning about whether formulas are true or false, our logic reasons about temporal landscapes. A temporal landscape may be thought of as representing the set of durations over which a statement is true. To help understand the practical implications of this approach, we give a wide variety of examples where this logic is used to reason about autonomous agents.

math.LO↗
arXiv2022

Fast Left Kan Extensions Using The Chase

We show how computation of left Kan extensions can be reduced to computation of free models of cartesian (finite-limit) theories. We discuss how the standard and parallel chase compute weakly free models of regular theories and free models of cartesian theories, and compare the concept of "free model" with a similar concept from database theory known as "universal model". We prove that, as algorithms for computing finite free models of cartesian theories, the standard and parallel chase are complete under fairness assumptions. Finally, we describe an optimized implementation of the parallel chase specialized to left Kan extensions that achieves an order of magnitude improvement in our performance benchmarks compared to the next fastest left Kan extension algorithm we are aware of.

cs.DB↗
arXiv2022

Generalized Lens Categories via functors $\mathcal{C}^{\rm op}\to\mathsf{Cat}$

Lenses have a rich history and have recently received a great deal of attention from applied category theorists. We generalize the notion of lens by defining a category $\mathsf{Lens}_F$ for any category $\mathcal{C}$ and functor $F\colon \mathcal{C}^{\rm op}\to\mathsf{Cat}$, using a variant of the Grothendieck construction. All of the mathematics in this note is straightforward; the purpose is simply to see lenses in a broader context where some closely-related examples, such as ringed spaces and open continuous dynamical systems, can be included.

math.CT↗
arXiv2022

Collectives: Compositional protocols for contributions and returns

We introduce a concept called a collective: an interface with a protocol for aggregating contributions and distributing returns. Through such a protocol, many members may participate in a mutual endeavor. We present a variety of real-world examples of collectives to explore the many kinds of things that members can contribute (such as time, work, ideas, or resources) as well as the many ways these contributions can be aggregated and returns distributed. In addition, we illustrate several ways in which new collectives can be constructed from old, alluding to the fact that all these constructions have a natural mathematical description within the category of polynomial functors equipped with a certain monoidal structure.

math.CT↗
arXiv2021

Deep neural networks as nested dynamical systems

There is an analogy that is often made between deep neural networks and actual brains, suggested by the nomenclature itself: the "neurons" in deep neural networks should correspond to neurons (or nerve cells, to avoid confusion) in the brain. We claim, however, that this analogy doesn't even type check: it is structurally flawed. In agreement with the slightly glib summary of Hebbian learning as "cells that fire together wire together", this article makes the case that the analogy should be different. Since the "neurons" in deep neural networks are managing the changing weights, they are more akin to the synapses in the brain; instead, it is the wires in deep neural networks that are more like nerve cells, in that they are what cause the information to flow. An intuition that nerve cells seem like more than mere wires is exactly right, and is justified by a precise category-theoretic analogy which we will explore in this article. Throughout, we will continue to highlight the error in equating artificial neurons with nerve cells by leaving "neuron" in quotes or by calling them artificial neurons. We will first explain how to view deep neural networks as nested dynamical systems with a very restricted sort of interaction pattern, and then explain a more general sort of interaction for dynamical systems that is useful throughout engineering, but which fails to adapt to changing circumstances. As mentioned, an analogy is then forced upon us by the mathematical formalism in which they are both embedded. We call the resulting encompassing generalization deeply interacting learning systems: they have complex interaction as in control theory, but adaptation to circumstances as in deep neural networks.

cs.LG↗
arXiv2021

Dirichlet polynomials and entropy

A Dirichlet polynomial $d$ in one variable ${\mathcal{y}}$ is a function of the form $d({\mathcal{y}})=a_n n^{\mathcal{y}}+\cdots+a_22^{\mathcal{y}}+a_11^{\mathcal{y}}+a_00^{\mathcal{y}}$ for some $n,a_0,\ldots,a_n\in\mathbb{N}$. We will show how to think of a Dirichlet polynomial as a set-theoretic bundle, and thus as an empirical distribution. We can then consider the Shannon entropy $H(d)$ of the corresponding probability distribution, and we define its length (or, classically, its perplexity) by $L(d)=2^{H(d)}$. On the other hand, we will define a rig homomorphism $h\colon\mathsf{Dir}\to\mathsf{Rect}$ from the rig of Dirichlet polynomials to the so-called rectangle rig, whose underlying set is $\mathbb{R}_{\geq0}\times\mathbb{R}_{\geq0}$ and whose additive structure involves the weighted geometric mean; we write $h(d)=(A(d),W(d))$, and call the two components area and width (respectively). The main result of this paper is the following: the rectangle-area formula $A(d)=L(d)W(d)$ holds for any Dirichlet polynomial $d$. In other words, the entropy of an empirical distribution can be calculated entirely in terms of the homomorphism $h$ applied to its corresponding Dirichlet polynomial. We also show that similar results hold for the cross entropy.

cs.IT↗
arXiv2021

Monitoring and Diagnosability of Perception Systems

Perception is a critical component of high-integrity applications of robotics and autonomous systems, such as self-driving vehicles. In these applications, failure of perception systems may put human life at risk, and a broad adoption of these technologies requires the development of methodologies to guarantee and monitor safe operation. Despite the paramount importance of perception systems, currently there is no formal approach for system-level monitoring. In this work, we propose a mathematical model for runtime monitoring and fault detection and identification in perception systems. Towards this goal, we draw connections with the literature on diagnosability in multiprocessor systems, and generalize it to account for modules with heterogeneous outputs that interact over time. The resulting temporal diagnostic graphs (i) provide a framework to reason over the consistency of perception outputs -- across modules and over time -- thus enabling fault detection, (ii) allow us to establish formal guarantees on the maximum number of faults that can be uniquely identified in a given perception system, and (iii) enable the design of efficient algorithms for fault identification. We demonstrate our monitoring system, dubbed PerSyS, in realistic simulations using the LGSVL self-driving simulator and the Apollo Auto autonomy software stack, and show that PerSyS is able to detect failures in challenging scenarios (including scenarios that have caused self-driving car accidents in recent years), and is able to correctly identify faults while entailing a minimal computation overhead (< 5 ms on a single-core CPU).

cs.RO↗
arXiv2021

Regular Calculi I: Graphical Regular Logic

What is ergonomic syntax for relations? In this first paper in a series of two, to answer the question we define regular calculi: a suitably structured functor from a category representing the syntax of regular logic to the category of posets, that takes each object to the poset of relations on that type. We introduce two major classes of examples, regular calculi corresponding to regular theories, and regular calculi corresponding to regular categories. For working in regular calculi, we present a graphical framework which takes as primitive the various moves of regular logic. Our main theorem for regular calculi is a syntax-semantics $2$-dimensional adjunction to regular categories.

math.CT↗
arXiv2021

Behavioral Mereology: A Modal Logic for Passing Constraints

Mereology is the study of parts and the relationships that hold between them. We introduce a behavioral approach to mereology, in which systems and their parts are known only by the types of behavior they can exhibit. Our discussion is formally topos-theoretic, and agnostic to the topos, providing maximal generality; however, by using only its internal logic we can hide the details and readers may assume a completely elementary set-theoretic discussion. We consider the relationship between various parts of a whole in terms of how behavioral constraints are passed between them, and give an inter-modal logic that generalizes the usual alethic modalities in the setting of symmetric accessibility.

cs.LO↗
arXiv2021

Wiring diagrams as normal forms for computing in symmetric monoidal categories

Applications of category theory often involve symmetric monoidal categories (SMCs), in which abstract processes or operations can be composed in series and parallel. However, in 2020 there remains a dearth of computational tools for working with SMCs. We present an "unbiased" approach to implementing symmetric monoidal categories, based on an operad of directed, acyclic wiring diagrams. Because the interchange law and other laws of a SMC hold identically in a wiring diagram, no rewrite rules are needed to compare diagrams. We discuss the mathematics of the operad of wiring diagrams, as well as its implementation in the software package Catlab.

cs.LO↗
arXiv2021

Proceedings of the 3rd Annual International Applied Category Theory Conference 2020

The third annual International Applied Category Theory Conference (ACT2020) was planned to take place at MIT in Cambridge, Massachusetts USA. However, the global COVID-19 pandemic made the prospect of holding a large in-person meeting impossible, and the event was thus held completely online. Holding the talks online had the new benefits of reducing carbon footprint, being inclusive of people from more parts of the world, and producing higher-quality video talks, which have been posted online for posterity. The ACT2020 contributions spanned a broad spectrum of application areas, including databases, dynamical systems, functional programming, game theory, lenses, neuroscience, probabilistic programming, natural language processing, quantum mechanics, and cyberphysical systems. Papers featured a broad range of categorical techniques. Papers in this Proceedings volume represents about half of the talks presented at ACT2020. Being included in the proceedings vs. not is not an indication of talk quality, but instead almost exclusively the choice of the authors, e.g. to present work already published elsewhere.

cs.DM↗
arXiv2020

Monitoring and Diagnosability of Perception Systems

Perception is a critical component of high-integrity applications of robotics and autonomous systems, such as self-driving cars. In these applications, failure of perception systems may put human life at risk, and a broad adoption of these technologies relies on the development of methodologies to guarantee and monitor safe operation as well as detect and mitigate failures. Despite the paramount importance of perception systems, currently there is no formal approach for system-level monitoring. In this work, we propose a mathematical model for runtime monitoring and fault detection of perception systems. Towards this goal, we draw connections with the literature on self-diagnosability for multiprocessor systems, and generalize it to (i) account for modules with heterogeneous outputs, and (ii) add a temporal dimension to the problem, which is crucial to model realistic perception systems where modules interact over time. This contribution results in a graph-theoretic approach that, given a perception system, is able to detect faults at runtime and allows computing an upper-bound on the number of faulty modules that can be detected. Our second contribution is to show that the proposed monitoring approach can be elegantly described with the language of topos theory, which allows formulating diagnosability over arbitrary time intervals.

cs.RO↗
arXiv2020

Dirichlet Polynomials form a Topos

One can think of power series or polynomials in one variable, such as $P(x)=2x^3+x+5$, as functors from the category $\mathsf{Set}$ of sets to itself; these are known as polynomial functors. Denote by $\mathsf{Poly}_{\mathsf{Set}}$ the category of polynomial functors on $\mathsf{Set}$ and natural transformations between them. The constants $0,1$ and operations $+,\times$ that occur in $P(x)$ are actually the initial and terminal objects and the coproduct and product in $\mathsf{Poly}_{\mathsf{Set}}$. Just as the polynomial functors on $\mathsf{Set}$ are the copresheaves that can be written as sums of representables, one can express any Dirichlet series, e.g.\ $\sum_{n=0}^\infty n^x$, as a coproduct of representable presheaves. A Dirichlet polynomial is a finite Dirichlet series, that is, a finite sum of representables $n^x$. We discuss how both polynomial functors and their Dirichlet analogues can be understood in terms of bundles, and go on to prove that the category of Dirichlet polynomials is an elementary topos.

math.CT↗