Banach Mazur Computable
Banach-Mazur computability investigates the limits of computation within specific frameworks, particularly focusing on the feasibility of solving optimization and classification problems using digital computers modeled as Turing machines. Current research explores these limitations in the context of deep learning, large language models, and other AI algorithms, examining the computability of optimizers and the inherent difficulty of tasks like robust PAC learning. These studies reveal fundamental constraints on algorithmic solvability and accuracy, impacting the design and trustworthiness of AI systems across diverse applications, from network programming to scientific modeling.
Papers
October 8, 2024
September 11, 2024
August 12, 2024
August 1, 2024
June 14, 2024
December 8, 2023
July 28, 2023
April 4, 2023
February 8, 2023
January 15, 2023
May 20, 2022
February 28, 2022