"Data Sketching via the Levy-Khintchine Lens" – Seth Pettie, Talks at TTIC

“Looking at Data Sketching Through the Levy-Khintchine Lens” Seth Pettie, University of Michigan Originally recorded on March 27, 2026, at TTIC. In this talk, Seth Pettie explores fundamental questions in data sketching, focusing on which functions can be efficiently estimated or sampled using polylogarithmic space. He introduces a novel perspective based on Levy processes and the Levy-Khintchine representation theorem, providing new insights into tractability and the possibility of universal sketching methods. Timestamps: 00:00 Introduction 01:45 Talk begins 58:10 Q&A #DataSketching #Algorithms #Theory #MachineLearning #StreamingAlgorithms #AI #LevyProcesses #Research #TTIC