Výsledky vyhledávání - "Abbott, Vincent P."

Report

FlashAttention on a Napkin: A Diagrammatic Approach to Deep Learning IO-Awareness

Optimizing deep learning algorithms currently requires slow, manual derivation, potentially leaving much performance untapped. Methods like FlashAttention have achieved a x6 performance improvement over native PyTorch by avoiding unnecessary data tra

Externí odkaz: http://arxiv.org/abs/2412.03317

Zobrazit plný text záznamu

Report

Diagrammatic Negative Information

Autor: Abbott, Vincent, Zardini, Gioele

The flow of information through a complex system can be readily understood with category theory. However, negative information (e.g., what is not possible) does not have an immediately evident categorical representation. The formalization of nategori

Externí odkaz: http://arxiv.org/abs/2404.03224

Zobrazit plný text záznamu

Report

Functor String Diagrams: A Novel Approach to Flexible Diagrams for Applied Category Theory

Autor: Abbott, Vincent, Zardini, Gioele

The study of abstraction and composition - the focus of category theory - naturally leads to sophisticated diagrams which can encode complex algebraic semantics. Consequently, these diagrams facilitate a clearer visual comprehension of diverse theore

Externí odkaz: http://arxiv.org/abs/2404.00249

Zobrazit plný text záznamu

Report

Neural Circuit Diagrams: Robust Diagrams for the Communication, Implementation, and Analysis of Deep Learning Architectures

Autor: Abbott, Vincent

Publikováno v: Transactions on Machine Learning Research (2024)

Diagrams matter. Unfortunately, the deep learning community has no standard method for diagramming architectures. The current combination of linear algebra notation and ad-hoc diagrams fails to offer the necessary precision to understand architecture

Externí odkaz: http://arxiv.org/abs/2402.05424

Zobrazit plný text záznamu