报告一
Title: 无监督深度学习及物理属性的融合
Speaker: 戚海蓉，美国田纳西大学电气工程与计算机科学系教授，IEEE Fellow，冈萨雷斯冠名教授
Time: 2019年7月10日上午9:30-10:30
Place: 计算机科学与技术学院320会议室
Abstract
Deep learning models have achieved tremendous success but often need large amount of training data which might not be readily available. This talk explores unsupervised learning and its application to hyperspectral image superresolution. Also in this example, we study ways to incorporate physical constraints to the network design in a natural fashion, such that the representations/outputs from the network can follow certain distribution and thus satisfy certain important physical constraints.
Speaker’s Biography
戚海蓉，美国田纳西大学电气工程与计算机科学系教授，IEEE Fellow，冈萨雷斯冠名教授。1992年和1995年分别获北京交通大学计算机系学士和硕士学位；1999年获北卡罗莱纳州立大学计算机工程系获博士学位。研究领域包括图像处理、协同信息处理、超光谱图像分析、计算机视觉和机器学习。获得了多项美国NSF，DARPA，IARPA，和NASA等项目资助。戚海蓉教授已在高水平期刊和推荐会议上发表论文200余篇，出版了2本关于机器视觉的书籍。文章谷歌学术引用次数8000余次，h-index是42。担任本领域国际期刊IEEE Transactions on Image Processing (TIP)编委。曾获NSF CAREER Award，荣获ICPR 2006、ICDCS 2009、WHISPERS 2015最佳论文奖，2012年获IEEE地球科学和遥感协会最有影响力论文奖。
报告二
Title: 面向联邦学习的用户级隐私攻击
Speaker: 王志波，博士，武汉大学国家网络安全学院教授
Time: 2019年7月10日上午10:30-11:30
Place: 计算机科学与技术学院320会议室
Abstract
联邦学习是一种分布式机器学习框架，近年来在隐私安全与机器学习领域受到广泛关注和研究。相比于传统的集中式学习框架，联邦学习将模型的训练过程转移到了用户端，仅需要用户周期性地提交模型参数更新就能完成模型训练，避免了服务端对用户数据的恶意访问和滥用。本次报告研究了联邦学习中的隐私问题，提出了一个基于恶意服务端的用户隐私数据重建攻击方法，通过建立一个多任务生成对抗网络模型来模拟用户的数据分布，并利用用户参数更新来计算其数据表征以重建特定用户隐私数据。相比已有的攻击方法只能重建表征某个类别的样本数据，我们的攻击方法可以实现用户级的数据重建，并通过手写数字分类和人脸识别两个任务验证了攻击有效性，阐明了模型参数更新中包含了过多的隐私信息，现有联邦学习框架仍存在安全隐患。
Speaker’s Biography
王志波，博士，武汉大学国家网络安全学院教授，入选湖北省“楚天学者”、武汉大学“珞珈青年学者”，以及荣获了ACM武汉学术新星奖。2007年毕业于浙江大学信息学院自动化专业，获学士学位；2014年毕业于美国田纳西大学，获计算机工程博士学位。研究方向包括物联网、移动感知与计算、网络安全与隐私保护、人工智能安全。在网络与安全领域著名期刊和会议上发表论文70余篇，其中CCF A类长文16篇，发表在TMC、TDSC、TPDS、ACM CCS、IEEE INFOCOM等顶级期刊和会议上，5篇论文入选ESI高被引论文。主持与参与多项国家级省部级项目，受邀担任IEEE ACCESS、KSII Transactions on Internet and Information Systems的期刊编委，IEEE INFOCOM、IEEE IPCCC、Globecom、ICC等多个国际会议的大会程序委员。现为IEEE高级会员、ACM会员及CCF会员，CCF物联网专委会委员，CCF网络与数据通信专委会委员，中国通信学会云计算与大数据应用委员会委员。
报告三
Title: Learn from Human Eyes & Brain for Handling Big Visual Data
Speaker: Weisi Lin, Nanyang Technological University
Time: 2019年7月10日下午3:30-4:30
Place: 计算机科学与技术学院320会议室
Abstract
Human eyes receive much more visual information than what can be processed by the related part of the brain in real time. In response to this big data challenge, the human visual system (HVS; consisting eyes, visual pathways, visual cortices, etc) has been evolved to equip with the effective visual attention (VA) mechanism to select the most informative or interesting region in a scene for focus at a time; for instance, we pay attention to a red flower among many green leaves in a picture, or a green leaf among many red flowers. The VA and its computational models have attracted continuous research effort since William James' time. It is beneficial to understand the VA and incorporate an appropriate model computationally in visual/multimedia signal processing, toward various AI-oriented tasks.
In this talk, we will first introduce the research problems associated with VA, as well as the relevant physiological and psychological ground. Afterward, we are to discuss the principle of computational VA modelling and the advances in the area, including approaches with handcrafted and deep-learnt features. Meaningful applications in perceptual quality assessment, image retargeting, video coding, computer graphics, and fast target identification are then demonstrated. The talk will also present our opinions toward possible future research and development.
Speaker’s Biography
Weisi Lin is an active researcher in image processing, perception-based signal modelling and assessment, video compression, and multimedia communication systems. In the said areas, he has published 180+ international journal papers and 230+ international conference papers, 7 patents, 9 book chapters, 2 authored books and 3 edited books, as well as excellent track record in leading and delivering more than 10 major funded projects (with over S$7m research funding). He earned his Ph.D from King’s College, University of London. He had been the Lab Head, Visual Processing, in Institute for Infocomm Research (I2R). He is a Professor and the Programme Director (Special Projects) in School of Computer Science and Engineering, Nanyang Technological University, where he also served as the Associate Chair (Graduate Studies) in 2013-2014.
He is a Fellow of IEEE and IET, and an Honorary Fellow of Singapore Institute of Engineering Technologists. He has been elected as a Distinguished Lecturer in both IEEE Circuits and Systems Society (2016-17) and Asia-Pacific Signal and Information Processing Association (2012-13), and given keynote/invited/tutorial/panel talks to 20+ international conferences during the past 10 years. He has been an Associate Editor for IEEE Trans. on Image Processing, IEEE Trans. on Circuits and Systems for Video Technology, IEEE Trans. on Multimedia, IEEE Signal Processing Letters, Quality and User Experience, and Journal of Visual Communication and Image Representation. He was also the Guest Editor for 7 special issues in international journals, and chaired the IEEE MMTC QoE Interest Group (2012-2014); he has been a Technical Program Chair for IEEE Int’l Conf. Multimedia and Expo (ICME 2013), International Workshop on Quality of Multimedia Experience (QoMEX 2014), International Packet Video Workshop (PV 2015), Pacific-Rim Conf. on Multimedia (PCM 2012) and IEEE Visual Communications and Image Processing (VCIP 2017). He believes that good theory is practical, and has delivered 10 major systems and modules for industrial deployment with the technology developed.
