Composable Uncertainty in Symmetric Monoidal Categories for Design Problems
Marius Furter, Yujun Huang, Gioele Zardini · 2025
Read Composable Uncertainty in Symmetric Monoidal Categories for Design Problems by Marius Furter, Yujun Huang, Gioele Zardini (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-17274.txt.
Introduction
Marius Furter, Yujun Huang, and Gioele Zardini bring uncertainty into a category theory framework for engineering design. Their 2025 paper starts from design problems modeled as open systems that can be wired together. Real systems, though, are never fully known, so the math must carry doubt as a first-class citizen.
Their construction swaps the arrows of the design category for maps valued in a Markov category, a home for uncertain processes. Different choices recover different flavors of uncertainty, from tunable parameters to probability spreads. The original clean structures survive the move. Applied to design problems, the result supports optimization under doubt and Bayesian learning over designs, the authors show.
Catalog record
| CREATOR | Marius Furter, Yujun Huang, Gioele Zardini |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 288 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-17274
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Composable Uncertainty in Symmetric Monoidal Categories for Design Problems Marius Furter, Yujun Huang, Gioele Zardini Abstract Applied category theory often studies symmetric monoidal categories (SMCs) whose morphisms represent open systems. These structures naturally accommodate complex wiring patterns, leveraging (co)monoidal structures for splitting and merging wires, or compact closed structures for feedback. A key example is the compact closed SMC of design problems ( 𝖣𝖯 \mathsf{DP} ), which enables a compositional approach to co-design in engineering. However, in practice, the systems of interest may not be fully known. Recently, Markov categories have emerged as a powerful framework for modeling uncertain processes. In this work, we demonstrate how to integrate this perspective into the study of open systems while preserving consistency with the underlying SMC structure. To this end, we employ the change-of-base construction for enriched categories, replacing the morphisms of a symmetric monoidal 𝒱 \mathcal{V} -category 𝒞 \mathcal{C} with parametric maps A → 𝒞 ( X , Y ) A\to\mathcal{C}(X,Y) in a Markov category induced by a symmetric monoidal monad. This results in a symmetric monoidal 2-category N ∗ 𝒞 N_{*}\mathcal{C} with the same objects as 𝒞 \mathcal{C} and reparametrization 2-cells. By choosing different monads, we capture various types of uncertainty. The category underlying 𝒞 \mathcal{C} embeds into N ∗ 𝒞 N_{*}\mathcal{C} via a strict symmetric monoidal functor, allowing (co)monoidal and compact closed structures to be transferred. Applied to 𝖣𝖯 \mathsf{DP} , this construction leads to categories of practical relevance, such as parametrized design problems for optimization, and parametrized distributions of design problems for decision theory and Bayesian learning. 1 Introduction The design of complex systems is a critical but challenging task for society. Difficulties arise from interactions between heterogeneous components, competing design objectives, diverse stakeholders with varied interests, and computationally intensive models. Co-design tackles these challenges by leveraging a categorical model of design problems called \mathsf{DP} [ 23 ] . \mathsf{DP} is a compact closed symmetric monoidal category (SMC), whose objects are partially ordered sets (posets), and whose morphisms are monotone maps from \mathcal{F}^{\text{op}}\times\mathcal{R} to the poset of truth values \mathsf{Bool}=\{\bot\leq\top\} . It models heterogeneous systems as processes providing functionalities f\in\mathcal{F} in return for resources r\in\mathcal{R} . Monotonicity captures the intuition that, if a resource r suffices to provide functionality f , then it also suffices for any worse functionality f^{\prime}\leq f . Moreover, any better resource r^{\prime}\geq r should also suffice to provide f . Posets present transparent trade-offs between design objectives and provide a common ground for stakeholder discussions. Moreover, \mathsf{DP} enables functorial queries which decompose complex design problems into simpler ones that can be solved efficiently. Current co-design methods address uncertainty using upper and lower bounds [ 5 ] . This approach is robust and well-suited for safety-critical applications, ensuring system performance even in worst-case scenarios. Furthermore, it is effective for approximating and solving continuous co-design via discretization. However, this method is unable to provide quantitative measures of uncertainty, such as the probability that a system will meet functional requirements given specific resource constraints. Understanding such trade-off between functionality, resource usage, and success probability is crucial in many cases. Furthermore, design problems often involve parametric uncertainties linked to design choices. For instance, when designing soft-robot manipulators, the choice of material introduces the elastic modulus as a parameter that follows a distribution. Moreover, its optimal value differs between components: a lower elastic modulus benefits hardware design by lowering driving force requirements, while a higher modulus benefits controller design by improving resistance to disturbances. Such examples highlight the need to incorporate quantitative uncertainty into co-design in a way that can depend on design parameters. We propose a general scheme for adding parametrized uncertainty to any SMC. Importantly, this process preserves compact closed and (co)monoidal structure. Our approach supports any uncertainty semantics modeled by symmetric monoidal monads. This includes intervals, subsets and distributions, whose Kleisli categories form Markov categories. The parametrization enables decision-making and learning for any process represented by a diagram in the SMC. In more detail, given a \mathcal{V} -category \mathcal{C} , we replace its arrows \mathcal{C}(X,Y) with parametric maps A\to\mathsf{M}\mathcal{C}(X,Y) , where \mathsf{M}\colon\mathcal{V}\to\mathcal{V} is a symmetric monoidal monad. The strength \nabla\colon\mathsf{M}(-)\otimes\mathsf{M}(=)\to\mathsf{M}(-\otimes=) allows us to compose f\colon A\to\mathsf{M}\mathcal{C}(X,Y) and g\colon B\to\mathsf{M}\mathcal{C}(Y,Z) according to A\otimes B\overset{f\otimes g}{\longrightarrow}\mathsf{M}\mathcal{C}(X,Y)\otimes\mathsf{M}\mathcal{C}(Y,Z)\overset{\nabla}{\longrightarrow}\mathsf{M}(\mathcal{C}(X,Y)\otimes\mathcal{C}(Y,Z))\overset{\mathsf{M}\fatsemi_{\mathcal{C}}}{\longrightarrow}\mathsf{M}\mathcal{C}(X,Z). In addition, any \mathcal{V} -arrow \varphi\colon A^{\prime}\to\mathsf{M}A allows us to reparametrize f to A^{\prime}\to\mathsf{M}\mathcal{C}(X,Y) using Kleisli composition, providing 2-categorical structure. Furthermore, when \mathcal{C} is monoidal, we can define a monoidal product analogous to composition. Hence, our construction promises to be a monoidal 2-category. We prove this using change-of-base: Given a monoidal functor N\colon\mathcal{V}\to\mathcal{W} , we can convert a \mathcal{V} -category \mathcal{C} into a \mathcal{W} -category N_{*}\mathcal{C} by replacing its arrows by N\mathcal{C}(X,Y) . Moreover, when N is symmetric and \mathcal{C} (symmetric) monoidal, then so is N_{*}C . Finally, there is an inclusion of \mathcal{C}_{0}\to(N_{*}\mathcal{C})_{0} of underlying categories that strictly preserves the monoidal product and coherence maps. We view the proposed construction as a change-of-base along \mathsf{L}_{\mathsf{M}}\colon\mathcal{V}\to\mathsf{Kl}_{\mathsf{M}} and \mathsf{Kl}_{\mathsf{M}}/(-)\colon\mathsf{Kl}_{\mathsf{M}}\to\mathsf{cat} , where \mathsf{L}_{\mathsf{M}} is the identity-on-objects functor sending f\mapsto f\fatsemi\eta , and \mathsf{Kl}_{\mathsf{M}}/(-) maps X to the slice category \mathsf{Kl}_{\mathsf{M}}/X . However, \mathsf{Kl}_{\mathsf{M}}/X is only symmetric monoidal up to invertible 2-cells. Nonetheless, when \mathsf{Kl}_{\mathsf{M}} is strict (as can be assumed by strictification), change-of-base yields a (symmetric) monoidal 2-category where interchange (and naturality of symmetry) hold up to coherent reparametrization isomorphisms. Applied to design problems, this construction leads to categories of practical relevance. We demonstrate that \mathsf{DP} may be viewed as enriched in sets, posets, and measurable spaces. Hence, we can choose monads in any of these setting to get parametrized subsets, intervals, and distributions of design problems. In all cases, the compact closed SMC structure transfers to the resulting 2-category. The remainder of the paper is structured as follows. Section 2 introduces the required background on enriched categories and change-of-base. Section 3 describes uncertainty monads and the partial slice functor. It concludes with an explicit description of the proposed scheme in Section 3.3 . Finally, Section 4 describes categories of uncertain DPs and demonstrates their usefulness. Acknowledgments. MF thanks the Digital Society Initiative for funding his Ph.D., the Rudge (1948) and Nancy Allen Chair for his visit to LIDS at MIT, and SNF Grant 200021_227719 for travel costs. 2 Background This section introduces enriched categories and the change-of-base construction. Moreover, it explains that change-of-base is 2-functorial and induces a transfer of symmetric monoidal structure. The latter result originates from Geoff Cruttwell’s Ph.D. thesis [ 7 ] , which provides an exceptionally clear exposition. We assume familiarity with SMCs [ 20 ] , which we usually denote (\mathcal{V},\otimes,I) with coherence maps \alpha,\rho,\lambda , and \sigma . The key definitions are gathered in Appendix A . Many SMCs we use are Cartesian, including \mathsf{Set} (sets and functions), \mathsf{Pos} (posets and monotone functions), \mathsf{Top} (topological spaces and continuous functions), \mathsf{Meas} (measurable spaces and measurable functions), and \mathsf{cat} (small categories and functors). Our 2-categories will always be strict and we use f\fatsemi g:=g\circ f for composition in diagrammatic order. Main example 2.1 (Design problems) . Let \mathsf{Bool}:=\{\bot\leq\top\} be the poset of truth values. Given posets \mathcal{F} and \mathcal{R} , a feasibility relation \Phi\colon\mathcal{F}\to\mathcal{R} is a monotone map \mathcal{F}^{\text{op}}\times\mathcal{R}\to\mathsf{Bool} . Posets and feasibility relations assemble into a compact closed SMC \mathsf{DP} of design problems [ 12