主 讲 人：谢传龙
讲座时间：2023-12-28
讲座地点：T2 栋 306 室

哈尔滨工业大学（深圳）学术讲座

演讲人Speaker： 谢传龙

题目Title:  Enhancing the Power of Out-of-Distribution Detection via Sample-Aware Model Selection

时间Date：   2023 年 12 月 28 日       Time：10：30-12：00

地点Venue：   T2 栋 306 室

内容摘要Abstract： 

Many deep learning systems have achieved state-of-the-art recognition performance when the training and testing data are identically distributed. However, neural networks make high-confidence predictions even for inputs that are completely unrecognizable and outside the training distribution, leading to a significant decline in model prediction performance or even complete failure. Therefore, the detection of out-of-distribution (OoD) testing samples is of great significance for the safe deployment of deep learning models in open-world applications. In this work, we propose a novel framework for sample-aware model selection to enhance the effectiveness of OOD detection and identify distributional shifts of features at different levels. Our algorithm determines, for each test input, which pre-trained models and which layer of features are able to recognize the test input as an OOD sample.

个人简介(About the speaker):

谢传龙， 北京师范大学统计与数据科学研究中心特聘副研究员，硕士生导师。博士毕业于香港浸会大学，曾担任暨南大学助理研究员、华为诺亚方舟实验室研究员。主要研究领域为稳健泛化，深度学习，模型检验。担任ICML,NeurlPS, ICLR, CVPR, AAAI, AISTATS审稿人，发表多篇重要论文。