SSE Weekly Colloquium | 理工学院系列研讨会邀请

香港中文大学（深圳）理工学院致力于为全校师生打造一个进行科研学术交流的平台，促进各理工领域学术交流并探讨其前沿研究发展。自2021年9月开始，理工学院将在每周五下午三点至四点 TD102组织开展 “Weekly Colloquium” 系列讲座，将邀请来自理工学院的教授或是特邀校外知名学者开展学术研讨会。欢迎全体教职员工以及学生与会！

研讨会主要内容

10月15日，第五场Weekly Colloquium由理工学院张纵辉教授为大家呈现。以下为本次研讨会的主要内容：

#

研讨会题目

Decentralized Non-Convex Federated Learning

#

研讨会摘要

Motivated by large-scale signal processing and machine learning applications, distributed optimization methods have been widely studied in the past decade. The recent emphasis on data privacy and security further calls for distributed optimization algorithms that do not need the distributively acquired raw data to be pooled or exchanged, such as the federated learning (FL). Based on how the data is partitioned across the nodes, the distributed learning problems are often divided into three categories. The first category, which is also the most widely studied, is to learn from distributed samples (LDS), where each distributed node owns part of the data samples only, and each sample has complete feature information. We will review some of the most representative LDS algorithms, including the consensus gradient descent methods, primal-dual based methods, and gradient tracking methods, and see how they inspire recent development of FL algorithms. The second category is to learn from distributed features (LDF), where each distributed node owns all the data samples, but they know part of the feature information only. Examples include multi-view learning in imaging processing and multi-omics biomedical analysis. The last one, which is also the most challenging one, is to learn from hybrid data (LHD) where the nodes has neither complete samples nor full feature information. We will report our recent progress in LDF and LHD problems, as well as our efforts to improve robustness of learning algorithms under non-ideal message exchanges.

#

讲者信息

Tsung-Hui Chang received the B.S. degree in electrical engineering and the Ph.D. degree in communications engineering from the National Tsing Hua University (NTHU), Hsinchu, Taiwan, in 2003 and 2008, respectively. From 2012 to 2015, he was an Assistant Professor with the Department of Electronic and Computer Engineering, National Taiwan University of Science and Technology (NTUST), Taipei, Taiwan. In August 2015, Dr. Chang joined the School of Science and Engineering, The Chinese University of Hong Kong, Shenzhen, China, as an Assistant Professor, and since August 2018, as an Associate Professor. Prior to being a faculty member, Dr. Chang held research positions with NTHU, from 2008 to 2011, and the University of California, Davis, CA, USA, from 2011 to 2012. During his PhD study, Dr. Chang was a visiting student and research assistant in the University of Minnesota, Minneapolis, MN, USA. His research interests include signal processing and optimization problems in data communications, machine learning and big data analysis.

Dr. Chang is a Senior Member of IEEE, an Elected Member of IEEE SPS Signal Processing for Communications and Networking Technical Committee (SPCOM TC) (2020/01-), the Funding Chair of IEEE SPS Integrated Sensing and Communication Technical Working Group (ISAC TWG), and the elected Regional Director-at-Large of IEEE SPS for the Asian-Pacific Region (2022/01-). He received the Young Scholar Research Award of NTUST in 2014, IEEE ComSoc Asian-Pacific Outstanding Young Researcher Award in 2015, and the IEEE Signal Processing Society (SPS) Best Paper Award in 2018. He has served the editorial board for major SP journals, including an Associate Editor of IEEE TRANSACTIONS ON SIGNAL PROCESSING (2014/08-2018/12), IEEE TRANSACTIONS ON SIGNAL AND INFORMATION PROCESSING OVER NETWORKS (2015/01-2018/12), IEEE OPEN JOURNAL OF SIGNAL PROCESSING (2020/01-present), and a Senior Area Editor of IEEE TRANSACTIONS ON SIGNAL PROCESSING (2021/02-present).

理工学院简介

理工学院坐落于粤港澳大湾区核心城市–深圳。学院秉承香港中文大学的优秀人文传统和浓厚学术底蕴，致力于培养具有国际视野、中华传统和社会担当的创新性高层次理工人才。作为学校的始创学院之一，以发展的眼光，开设了5个本科专业（电子信息工程、数学与应用数学、新能源科学与工程、化学、金融工程），2个硕士项目（供应链与物流管理、通信工程）和3个博士项目（计算机与信息工程、材料科学与工程、数学）。凭借香港中文大学崇高的办学理念和完善的学术体系，在来自全球的优质师资团队的努力下，培育了一届又一届优秀毕业生。在全球突发疫情的复杂国际形势影响之下，我院毕业生依然保持很高的就业率。

本系列讲座面向全校师生，无需提前报名预约，欢迎各位积极参加，敬请期待每期内容！让我们齐聚TD102共同探讨理工科的奥秘！

The School of Science and Engineering is committed to creating a platform for students and faculty members to exchange ideas and explore cutting-edge research developments in various fields of science and engineering. Starting from 2021 fall, SSE will organize the "Weekly Colloquium" series of seminars every Friday from 3:00 to 4:00 pm in TD102. Each week, SSE will invite our faculty members and distinguished guest speakers from outside the University to conduct academic seminars.  All students and faculty members are invited to join the weekly colloquium! 

About the Colloquium

The speaker on October 15 is Prof. Tsung-Hui CHANG of SSE. The following is the detailed information about the seminar.

#

Topic

Decentralized Non-Convex Federated Learning

#

Abstract

Motivated by large-scale signal processing and machine learning applications, distributed optimization methods have been widely studied in the past decade. The recent emphasis on data privacy and security further calls for distributed optimization algorithms that do not need the distributively acquired raw data to be pooled or exchanged, such as the federated learning (FL). Based on how the data is partitioned across the nodes, the distributed learning problems are often divided into three categories. The first category, which is also the most widely studied, is to learn from distributed samples (LDS), where each distributed node owns part of the data samples only, and each sample has complete feature information. We will review some of the most representative LDS algorithms, including the consensus gradient descent methods, primal-dual based methods, and gradient tracking methods, and see how they inspire recent development of FL algorithms. The second category is to learn from distributed features (LDF), where each distributed node owns all the data samples, but they know part of the feature information only. Examples include multi-view learning in imaging processing and multi-omics biomedical analysis. The last one, which is also the most challenging one, is to learn from hybrid data (LHD) where the nodes has neither complete samples nor full feature information. We will report our recent progress in LDF and LHD problems, as well as our efforts to improve robustness of learning algorithms under non-ideal message exchanges.

#

The Speaker

Tsung-Hui Chang received the B.S. degree in electrical engineering and the Ph.D. degree in communications engineering from the National Tsing Hua University (NTHU), Hsinchu, Taiwan, in 2003 and 2008, respectively. From 2012 to 2015, he was an Assistant Professor with the Department of Electronic and Computer Engineering, National Taiwan University of Science and Technology (NTUST), Taipei, Taiwan. In August 2015, Dr. Chang joined the School of Science and Engineering, The Chinese University of Hong Kong, Shenzhen, China, as an Assistant Professor, and since August 2018, as an Associate Professor. Prior to being a faculty member, Dr. Chang held research positions with NTHU, from 2008 to 2011, and the University of California, Davis, CA, USA, from 2011 to 2012. During his PhD study, Dr. Chang was a visiting student and research assistant in the University of Minnesota, Minneapolis, MN, USA. His research interests include signal processing and optimization problems in data communications, machine learning and big data analysis.

Dr. Chang is a Senior Member of IEEE, an Elected Member of IEEE SPS Signal Processing for Communications and Networking Technical Committee (SPCOM TC) (2020/01-), the Funding Chair of IEEE SPS Integrated Sensing and Communication Technical Working Group (ISAC TWG), and the elected Regional Director-at-Large of IEEE SPS for the Asian-Pacific Region (2022/01-). He received the Young Scholar Research Award of NTUST in 2014, IEEE ComSoc Asian-Pacific Outstanding Young Researcher Award in 2015, and the IEEE Signal Processing Society (SPS) Best Paper Award in 2018. He has served the editorial board for major SP journals, including an Associate Editor of IEEE TRANSACTIONS ON SIGNAL PROCESSING (2014/08-2018/12), IEEE TRANSACTIONS ON SIGNAL AND INFORMATION PROCESSING OVER NETWORKS (2015/01-2018/12), IEEE OPEN JOURNAL OF SIGNAL PROCESSING (2020/01-present), and a Senior Area Editor of IEEE TRANSACTIONS ON SIGNAL PROCESSING (2021/02-present).

School Profile

School of Science and Engineering is located in Shenzhen, a core city of the Guangdong-Hong Kong-Macao Greater Bay Area. Adhering to the rich CUHK's academic tradition with high emphasis on humanity elements, the School is committed to nurturing innovative technology leaders embracing China and the world with global visions and social responsibility.

As one of the founding schools of the University, SSE has so far offered five undergraduate programmes (Electronic Information Engineering, Mathematics and Applied Mathematics, New Energy Science and Engineering, Chemistry, and Financial Engineering), two master programmes (Supply Chain and Logistics Management, and Communication Engineering), and four PhD programmes (Computer and Information Engineering, Materials Science and Engineering, Energy Science and Engineering, and Mathematics). Following the lofty educational philosophy and sound academic system of CUHK, SSE and its extraordinary faculty from all over the world have produced a succession of outstanding graduates. Even under the complex international situation of the global pandemic, SSE graduates still secured a high employment rate.

This series of seminars is open to all students and faculty members, and no pre-registration is required. Let's gather in TD102 to learn about the beauty of science and engineering!

点击以下链接，进入理工时刻：