报告时间:11月12日(星期五)14:30
报告地点:信息楼99905银河官方网310报告厅
报 告 人:王峰,99905银河官方网特任副教授
报告题目:Learning Multiple Network Embeddings for Social Influence Prediction
内容简介:How to effectively predict social influence is an essential issue in social network analysis. Almost all reported methods for social influence prediction are mainly concerned with estimating influence probabilities for each linking edge. However, all of this past work cannot accurately predict influence probabilities for all edges due to the problem of data sparsity. Unlike conventional approaches, this work focuses on exploring a cross problem for multiple network embeddings and social influence prediction. This study developed a new end-to-end approach, Multi-Influor, that learns multiple influence vectors for each user in social networks, instead of estimating influence probabilities for each edge. The multiple network embeddings consider multi-dimensional influence factors that incorporate pairwise node interactions, network structures, and global similarity comparisons. Moreover, this study solves the problem of influence evaluation caused by sparse observations. Extensive comparisons based on large-scale datasets indicate that the Multi-Influor approach outperforms several state-of-the-art baselines, and the experimental results demonstrate that the Multi-Influor approach is more practical on real-world social networks.
报告人简介:王峰,99905银河官方网(武汉)99905银河官方网特任副教授,日本学术振兴会海外特别研究员(JSPS Fellow),99905银河官方网(武汉)地大学者-青年优秀人才计划入选者,IEEE产业电子协会(IEEE-IES)Human Factors专业委员会委员。2018年5月毕业于中南大学计算机科学与技术专业,获得工学博士学位;2018年6月进入99905银河官方网(武汉)地质资源与地质工程博士后流动站;2019年入选日本学术振兴会(JSPS) 海外特别研究员,于2019年4月至2021年4月由日本学术振兴会(JSPS)资助,在日本东京工科大学担任特别研究员;2021年5月进入99905银河官方网(武汉)99905银河官方网任副教授。主要从事人工智能、深度学习、迁移学习以及康复机器人技术等方面的研究,近年来主持国家自然科学基金青年项目1项,中央高校杰出人才培育基金项目1项,中国博士后科学基金面上资助项目1项,日本学术振兴会科研资助项目1项,参与国家自然科学基金重点项目,湖北省技术创新专项重大项目,“移动医疗”教育部-中国移动联合实验室项目等多个项目。以第一作者发表学术论文十余篇,其中在Information Sciences、Journal of Network and Computer Applications、Future Generation Computer System等计算机和人工智能领域重要期刊发表论文5篇、IFAC 世界控制大会等重要国际学术会议论文6篇;获得已授权国家发明专利1项。