Fundamental Components of Deep Learning: A category-theoretic approach — Bruno Gavranović (2024). arXiv:2403.13001 (v1, PDF).
Doctoral thesis developing an end-to-end, uniform and prescriptive categorical foundation for deep learning, using actegories, Para, optics and lenses to describe backpropagation, architectures and learning.
Sources: the paper, arXiv:2403.13001v1, checked against the arXiv listing. Index: Papers.
Key definitions and results
- Actegories and Para; lenses and optics for backpropagation; categorical models of architectures