![[Oxford Seminar] Xiaoyan Li | A novel modular and compositional framework for System Dynamics @ToposInstitute](https://i.ytimg.com/vi/2zS8Kwjavyo/maxresdefault.jpg)
Uploaded March 2025 | Updated September 2026, 2 weeks ago
Oxford Seminar, 27th of February 2025
Title: A next-generation modular and compositional framework for System Dynamics modeling and beyond
System Dynamics (SD) is a powerful methodology for understanding and managing complex systems over time, with applications in various fields such as public health, business, and environmental management. Despite its strengths, traditional System Dynamics modeling methods face several critical limitations, including a lack of modularity, inadequate representation of complex relationships, stratification that obscures model transparency, rigid coupling of model syntax and semantics, and limited support for model composition and reuse. These challenges restrict the scalability, adaptability, and accuracy of models, particularly when applied to large-scale or interdisciplinary systems.
To address these limitations, this presentation introduces a novel framework grounded in Applied Category Theory (ACT), which provides a rigorous mathematical foundation for representing, relating, composing, and stratifying complex systems. By leveraging ACT, this research establishes a next-generation SD modeling approach that integrates mathematical rigor with practical utility, significantly enhancing the flexibility, expressiveness, and applicability of System Dynamics for researchers and practitioners across various domains.
To contextualize System Dynamics modeling in the realm of infectious disease simulation, the presentation will begin with an overview of a COVID-19 theory-based machine learning model, developed incorporating the Bayesian methodology of Particle Filtering algorithm to a Covid-19 System Dynamics model. This model played a pivotal role in supporting decision-making across 17 Canadian jurisdictions during the COVID-19 pandemic by providing daily transmission monitoring, short-term projections, and counterfactual intervention analysis over a year.
Oxford Seminar, 27th of February 2025
Title: A next-generation modular and compositional framework for System Dynamics modeling and beyond
System Dynamics (SD) is a powerful methodology for understanding and managing complex systems over time, with applications in various fields such as public health, business, and environmental management. Despite its strengths, traditional System Dynamics modeling methods face several critical limitations, including a lack of modularity, inadequate representation of complex relationships, stratification that obscures model transparency, rigid coupling of model syntax and semantics, and limited support for model composition and reuse. These challenges restrict the scalability, adaptability, and accuracy of models, particularly when applied to large-scale or interdisciplinary systems.
To address these limitations, this presentation introduces a novel framework grounded in Applied Category Theory (ACT), which provides a rigorous mathematical foundation for representing, relating, composing, and stratifying complex systems. By leveraging ACT, this research establishes a next-generation SD modeling approach that integrates mathematical rigor with practical utility, significantly enhancing the flexibility, expressiveness, and applicability of System Dynamics for researchers and practitioners across various domains.
To contextualize System Dynamics modeling in the realm of infectious disease simulation, the presentation will begin with an overview of a COVID-19 theory-based machine learning model, developed incorporating the Bayesian methodology of Particle Filtering algorithm to a Covid-19 System Dynamics model. This model played a pivotal role in supporting decision-making across 17 Canadian jurisdictions during the COVID-19 pandemic by providing daily transmission monitoring, short-term projections, and counterfactual intervention analysis over a year.
![[Oxford Seminar] Xiaoyan Li | A novel modular and compositional framework for System Dynamics](https://i.ytimg.com/vi/2zS8Kwjavyo/mqdefault.jpg "[Oxford Seminar] Xiaoyan Li | A novel modular and compositional framework for System Dynamics
Oxford Seminar, 27th of February 2025
Title: A next-generation modular and compositional framework for System Dynamics modeling and beyond
System Dynamics (SD) is a powerful methodology for understanding and managing complex systems over time, with applications in various fields such as public health, business, and environmental management. Despite its strengths, traditional System Dynamics modeling methods face several critical limitations, including a lack of modularity, inadequate representation of complex relationships, stratification that obscures model transparency, rigid coupling of model syntax and semantics, and limited support for model composition and reuse. These challenges restrict the scalability, adaptability, and accuracy of models, particularly when applied to large-scale or interdisciplinary systems.
To address these limitations, this presentation introduces a novel framework grounded in Applied Category Theory (ACT), which provides a rigorous mathematical foundation for representing, relating, composing, and stratifying complex systems. By leveraging ACT, this research establishes a next-generation SD modeling approach that integrates mathematical rigor with practical utility, significantly enhancing the flexibility, expressiveness, and applicability of System Dynamics for researchers and practitioners across various domains.
To contextualize System Dynamics modeling in the realm of infectious disease simulation, the presentation will begin with an overview of a COVID-19 theory-based machine learning model, developed incorporating the Bayesian methodology of Particle Filtering algorithm to a Covid-19 System Dynamics model. This model played a pivotal role in supporting decision-making across 17 Canadian jurisdictions during the COVID-19 pandemic by providing daily transmission monitoring, short-term projections, and counterfactual intervention analysis over a year.")
![[Berkeley Seminar] William Troiani | Programs as Singularities](https://i.ytimg.com/vi/3DwkYrleuLU/mqdefault.jpg "[Berkeley Seminar] William Troiani | Programs as Singularities
Speaker: William Troiani (https://williamtroiani.github.io/)
Title: Programs as Singularities: differentiable computation and the geometry of Bayesian model selection
Abstract: I will describe a setting in which the internal structure of an algorithm becomes visible to Bayesian statistics. Using differentiable linear logic and its vector-space semantics, ordinary Turing-machine codes can be embedded into a smooth space of noisy codes, where perturbations correspond to read errors on the description tape of a universal Turing machine. This turns a program into a point in the parameter space of a statistical model, whose local geometry records how its output probabilities change under perturbation. I will explain how singular learning theory enters through local Bayesian evidence: when comparing neighbourhoods of candidate implementations, the asymptotics are controlled by the local learning coefficient. The main theorem identifies Taylor coefficients of the induced error/loss germ with weighted counts of error syndromes in the underlying computation. As a consequence, runtime error correction forces high-order flatness of the associated statistical singularity and yields bounds on the local learning coefficient. I will close by discussing this as an example of structural Bayesianism: the idea that Bayesian inference can be sensitive not only to what a model predicts, but also to how those predictions are algorithmically implemented.
Date: May 12, 2026")
![[2-torial] David tells Joanna about algebraic patterns](https://i.ytimg.com/vi/3VP3ZfgfpTg/mqdefault.jpg "[2-torial] David tells Joanna about algebraic patterns
[2-torial] David tells Joanna about algebraic patterns")
![[Berkeley Seminar] Evan Patterson | How to Prove Equations Using Diagrams](https://i.ytimg.com/vi/3wLDkjocE2c/mqdefault.jpg "[Berkeley Seminar] Evan Patterson | How to Prove Equations Using Diagrams
Title: How to Prove Equations Using Diagrams: A Category Theorist Looks at E-Graphs
Abstract: First invented in 1980, e-graphs are receiving new attention as a powerful and adaptable data structure to keep the books for equational reasoning. In this talk on work in progress, we explore an operational semantics for e-graphs based on category-theory. We argue that classic concepts from category theory, particularly diagrams and initial functors, provide a conceptual foundation for the diagrammatic reasoning performed in e-graphs.
https://topos.institute/blog/2025-05-27-e-graphs-1/
Date: 2025-05-20
https://topos.institute/events/berkeley-seminar/")
![[Berkeley Seminar] David Jaz Myers | Categorical Algebra with Segal Conditions](https://i.ytimg.com/vi/52AsVTxHXU0/mqdefault.jpg "[Berkeley Seminar] David Jaz Myers | Categorical Algebra with Segal Conditions
Title: Categorical Algebra with Segal Conditions
Abstract: There are many ways to present algebraic structures categorically: monads, Lawvere theories, limit sketches, and more. In this talk, well learn about Yet Another Way to Present Algebra (YAWPA): Segal conditions and Chu and Haugsengs algebraic patterns. The basic idea behind algebraic patterns is that in many cases of categorical algebra we have a notion of *pasting diagram* which tells us how we are going to compose things. These pasting diagrams are made out of elementary features things like boxes and wires and can swallow other diagrams whole by putting subdiagrams into boxes. These two kinds of morphisms, diagram inclusions and swallowings, organize into a factorization system on a category of pasting diagrams. Presheaves on this category of diagrams which send any pasting diagram to the limit over all its elementary subdiagrams is an algebra composing according to those pasting diagrams; that single axiom is the so-called Segal condition. Well see how this works for categories and double categories in particular, and, time-permitting, see how to derive the notion of virtual double category from the algebraic pattern for categories and muse about the appropriate virtual triple categories which are similarly derived from the algebraic pattern for double categories.
https://topos.institute/events/berkeley-seminar/")
![[Oxford Seminar] Olga Paris-Romanskevich | Gender and science: what every mathematician should know](https://i.ytimg.com/vi/5Zmlp2bmhmI/mqdefault.jpg "[Oxford Seminar] Olga Paris-Romanskevich | Gender and science: what every mathematician should know
Oxford Seminar, April 30 2026
Speaker: Olga Paris-Romanskevich
Full Title: Which results on gender and science every mathematician should
know
Abstract: Based on a selection of works in social psychology, sociology and cultural studies, I will discuss several epistemological approaches of how one can think the exclusion of women and gender minorities from sciences and mathematics in particular. In the discussion following the talk, we will collectively approach what can be done, at the individual and collective level, in order to prevent violence and exclusion from happening.")
![[Oxford Seminar] Owen Lynch | Markov Semigroups](https://i.ytimg.com/vi/5bBPz4GpBIw/mqdefault.jpg "[Oxford Seminar] Owen Lynch | Markov Semigroups
Oxford Seminar, 14th of February 2025")
![[DOTS Lectures] 15. Representability of reachability and double operad algebras](https://i.ytimg.com/vi/6UGVkOVvM5U/mqdefault.jpg "[DOTS Lectures] 15. Representability of reachability and double operad algebras
Part of a lecture series on the Double Operadic Theory of Systems (DOTS) presented by David Jaz Myers.
Some material from these lectures can be found in Davids book on categorical systems theory:
https://www.davidjaz.com/Papers/DynamicalBook.pdf")
![[2-torial] Jason tells Tim about Doctrinal Adjunctions](https://i.ytimg.com/vi/6oDQcA6dU8w/mqdefault.jpg "[2-torial] Jason tells Tim about Doctrinal Adjunctions
[2-torial] Jason tells Tim about Doctrinal Adjunctions")
![[Berkeley Seminar] B. Rousse | Heidegger, Skill, and the Conversational Structure of Human Work](https://i.ytimg.com/vi/6xwODsGadn4/mqdefault.jpg "[Berkeley Seminar] B. Rousse | Heidegger, Skill, and the Conversational Structure of Human Work
Title: Heidegger, Skill, and the Conversational Structure of Human Work
Abstract: AI is transforming how we think about work and intelligence. Many are claiming that AI systems will soon be able to do all the work that human beings do. This claim has a long and somewhat dubious history in the field of AI. To assess its plausibility, it helps to develop a deeper account of the nature of human work and agency. In this talk, I share work in progress aimed at this goal. I trace a tradition that runs from Heidegger’s ontology of human agency, to the theory of skilled activity articulated by Stuart and Hubert Dreyfus (the Dreyfus Skill Model), to the account of the conversational structure of human work developed by Terry Winograd and Fernando Flores in Understanding Computers and Cognition.
Date: 2025-04-29
https://topos.institute/events/berkeley-seminar/")
