세미나 및 초청강연

The generalization error of max-margin classifiers in the overparametrized regime

관리자l 2021-09-06l 조회수 382
일시 : 2021-10-08 11:00 ~ 12:00
연사 : 손영탁 (Department of Statistics, Stanford university)
장소 : ZOOM 활용 비대면 세미나
비고 : https://snu-ac-kr.zoom.us/j/89989108345

[  ]

Many modern learning methods, such as deep neural networks, are so complex that they perfectly fit the training data. Despite this, they generalize well to the unseen data. This talk will discuss an example of this phenomenon, namely the max-margin estimators for binary classification tasks, which achieves vanishing training error for separable data.
First, I will determine sharp asymptotics for the generalization error of max-margin classifiers for the Gaussian data. Then, I will show how we can use the result to study nonlinear random features model, two-layer neural networks with random first layer weights. Finally, I will emphasize several statistical insights which can be drawn from such mathematical analysis.


 첨부파일 (1개)