"Efficient Learning Algorithms under Contamination" – Konstantinos Stavropoulos,
Автор: TTIC
Загружено: 2025-11-24
Просмотров: 71
Описание:
“Efficient Learning Algorithms under (Heavy) Contamination”
Konstantinos Stavropoulos, University of Texas at Austin
Originally recorded on November 12, 2025, at TTIC.
In this talk, Konstantinos Stavropoulos presents new theoretical results in supervised learning from heavily contaminated datasets. He introduces outlier-removal algorithms inspired by distribution shift techniques and shows how low-degree polynomial approximators and sandwiching approximators enable efficient learning under bounded and even heavy contamination. The talk highlights advances in robust learning theory and includes new quasipolynomial-time results for learning AC⁰ circuits under contamination.
Timestamps:
00:00 Introduction
01:35 Talk begins
57:20 Q&A
#MachineLearning #Theory #RobustLearning #Contamination #Algorithms #AI #LearningTheory #AC0 #Research #YoungResearchers #TTIC
Повторяем попытку...
Доступные форматы для скачивания:
Скачать видео
-
Информация по загрузке:
