Towards a Compositional Framework for Convex Analysis (with Applications to Probability Theory)
Abstract: We introduce a compositional framework for convex analysis based on the notion of convex bifunction of Rockafellar. This framework is well-suited to graphical reasoning, and exhibits rich dualities such as the Legendre-Fenchel transform, while generalizing formalisms like graphical linear algebra, convex relations and convex programming. We connect our framework to probability theory by interpreting the Laplace approximation in its context: The exactness of this approximation on normal distributions means that logdensity is a functor from Gaussian probability (densities and integration) to concave bifunctions and maximization.
- Fujii, S.: A categorical approach to l-convexity. arXiv preprint arXiv:1904.08413 (2019)
- McCullagh, P.: Tensor methods in statistics. Courier Dover Publications (2018)
- Puhalskii, A.: Large deviations and idempotent probability. CRC Press (2001)
- Stein, D.: Structural foundations for probabilistic programming languages. University of Oxford (2021)
- Touchette, H.: Legendre-fenchel transforms in a nutshell. Unpublished Report (Queen Mary University of London) (2005)
- Willems, J.C.: Constrained probability. In: 2012 IEEE International Symposium on Information Theory Proceedings. pp. 1049–1053 (2012). https://doi.org/10.1109/ISIT.2012.6283011
- Willerton, S.: The Legendre-Fenchel transform from a category theoretic perspective. arXiv preprint arXiv:1501.03791 (2015)
Paper Prompts
Sign up for free to create and run prompts on this paper using GPT-5.
Top Community Prompts
Collections
Sign up for free to add this paper to one or more collections.