报告时间：12月13日（周三）下午14:00-16:00
报告地点：浙江大学紫金港校区蒙民伟楼402会议室
报告人：阳及 助理教授
 
个人简介：阳及博士现于埃默里大学计算机系任助理教授。他于2014年在浙江大学获计算机专业学士学位，并于2020年在伊利诺伊大学香槟分校获计算机专业博士学位。他的科研方向包括图数据挖掘，应用机器学习，知识图谱和联邦学习，以及它们在社交网络，推荐系统，神经科学和医疗保健等领域的应用。阳及博士的研究成果发表于横跨数据挖掘和健康信息学领域的120余篇顶会或顶刊论文。他也曾获得UIUC 2020博士论文奖，ICDM 2020最佳论文奖，KDD 2022健康日最佳论文奖，ML4H 2022最佳论文奖，Amazon Research Award，Microsoft Accelerating Foundation Models Research Award，OpenAI Research Access Award等荣誉。
 
题目1：Advance healthcare with multimodal structured knowledge
摘要：Health data is inherently multimodal and complicated, which necessitates domain expertise to interpret. Moreover, in the medical field, there is a crucial need for rigorous explanations rooted in domain knowledge. Such insights go beyond target predictions, fostering novel scientific discoveries. Toward these, my research vision focuses on advancing healthcare with multimodal structured knowledge. This involves transforming complicated and large-scale datasets into knowledge, which then serves as the foundation for developing interpretable AI models for downstream applications. Two themes from my previous work, namely brain network analysis (MICCAI'22 Oral, IEEE Transaction on Medical Imaging, KDD'23, NeurIPS'22 Spotlight) and multimodal data analysis (NeurIPS'23, ACL'23, ECIR'22, ECML'21), underscore this vision. Looking forward, I am passionately dedicated to refining the evaluation of large models, advancing data-centric AI, and harmonizing domain knowledge from medical professionals within healthcare. My long-term aspiration is to contribute to a health system that values interpretability, reliability, and equity for everyone.
 
题目2：Empower Graph Machine Learning for Brain Network Analysis
摘要：Brain imaging has become an important tool in modern neuroscience, helping us to better understand the functioning and structure of the human brain. Researchers are now using advanced machine learning techniques such as Convolutional Neural Networks (CNNs) and Graph Neural Networks (GNNs) to analyze brain networks extracted from imaging data. In this talk, we will introduce two works about exploring the potential of GNNs to refine the analysis of brain imaging data and provide reliable predictions. One work is FBNETGEN. It employs a unique graph generator to transform raw time-series features into task-oriented brain networks. Another research work is Brain Network Transformer, which studies the integration of Graph Transformer models specifically designed for analyzing brain data in the context of brain network analysis.