讲座题目：NeuroAI and its Applications in Model Compression

主讲人：香港城市大学范凤磊博士

讲座时间：2025年5月22日（周四）14：30-15：30

讲座地点：6号学院楼402会议室

主办单位：数据科学学院、浙江省2011“数据科学与大数据分析协同创新中心”

摘 要：

Deep learning, particularly through deep artificial neural networks, has emerged as a dominant force across numerous critical research domains over the past decade. While neural networks were originally conceived to emulate the human brain, contemporary advancements in deep learning have not been primarily fueled by our expanding knowledge of neuroscience. The human brain remains the most sophisticated intelligent system known to date, yet much of its functionality remains poorly understood. Despite this, it is evident that current deep learning systems still significantly lag behind the human brain in key aspects including computational efficiency, interpretability, and memory capacity.

Given the remarkable capabilities of biological neural systems, we posit that neuroscience can serve as both an inspirational framework and a validation mechanism for advancing artificial intelligence. In this presentation, we explore how principles derived from genomic bottleneck mechanisms can be innovatively applied to deep learning architectures. Our particular focus will be on addressing fundamental challenges in model compression - a crucial step toward enabling efficient deployment of large-scale neural networks in real-world applications. This interdisciplinary approach not only bridges neuroscience and artificial intelligence but also opens new avenues for developing more efficient and biologically plausible learning systems.

主讲人简介：

范凤磊博士现任香港城市大学数据科学系tenure-track助理教授，并担任华为重点项目首席科学家。他2017年本科毕业于哈尔滨工业大学，2021年于美国伦斯勒理工学院获得博士学位，师从国际知名医学影像学家王革教授。在博士期间，他荣获IBM AI Horizon Fellowship，并受邀在MIT-IBM AI Watson实验室实习。此后，他先后在康奈尔大学和香港中文大学从事博士后及研究助理教授工作，积累了丰富的跨学科研究经验。研究成果丰硕，已在《JMLR》、《CVPR》、《IEEE TNNLS》、《IEEE TMI》、《IEEE TCI》、《IEEE TCSVT》和《IEEE TAI》等人工智能与数据科学领域顶级期刊和会议发表论文20余篇。其博士论文荣获国际神经网络学会（INNS）2021年杰出博士论文奖，一篇研究论文入选2024年CVPR最佳论文奖候选（从超1万篇投稿中脱颖而出，仅26篇入选），另有一篇论文获IEEE TRPMS最佳论文奖及ESI高被引论文荣誉。此外，他多次在AAAI、WWW、IJCNN等顶级会议组织教学教程，受到学界广泛关注。

欢迎感兴趣的师生积极参加！