Quantitative Bound
Quantitative bounds research focuses on establishing rigorous mathematical limits on the performance and behavior of various models, primarily neural networks. Current efforts concentrate on deriving precise bounds for quantities like network gradients and Hessians, probabilities of successful outcomes (e.g., in robotic grasping), and approximation errors in solving partial differential equations. This work is crucial for improving the reliability and predictability of machine learning models and for providing theoretical guarantees for their application in safety-critical domains.
Papers
June 6, 2024
September 29, 2023
July 12, 2023
October 23, 2022
June 6, 2022
May 31, 2022