SSE Talk | 下周讲座预告

Dear all,

You are cordially invited to five academic seminars. Please find the details as follows.

讲座1

✦

✦

✦

✦

✦

讲座题目：低质量视觉内容增强与感知

日期&时间：2023年12月16日（周六）10:30-11:30

地点：教学楼D 103

主讲嘉宾：任文琦教授，中山大学

主持人：李镇教授，香港中文大学（深圳）

讲座语言：中文

摘要：

本次报告首先主要围绕图像视频常见的自然干扰和对抗干扰展开介绍。主要介绍几种图像视频恢复及增强的方法研究，针对成像过程中模糊，雨雾，低分辨率，低动态范围等不同质量退化的成因，提出针对性的解决方案。其次，介绍基于对抗噪音的低质量图像和伪装目标检测的方法，利用正向的对抗噪音对干净标签进行对抗攻击，获得与原始干净样本接近又兼顾下游目标检测效果的伪标签，用其监督训练得到的图像恢复模型可以在提升图像清晰度的同时提升目标检测准确度。  

讲者介绍：

任文琦，中山大学 “百人计划”副教授，主持国家优秀青年科学基金。天津大学与美国加州大学Merced分校联合培养博士。主要研究方向包括计算机视觉与网络空间内容安全。在本领域内发表CCF-A类国际期刊和会议学术论文70余篇，谷歌学术引用10000余次，6篇论文入选ESI高被引论文。担任CVPR、ICLR、AAAI等人工智能与计算机视觉领域多个国际会议的领域主席AC和高级程序委员会委员SPC，《中国图象图形学报》青年编委。主持国家基金3项，2项重点研发计划子课题，及多项腾讯、百度等企业项目，积极促进了研究成果产业化。研究成果应用于公安部专项行动、华为、腾讯等。获中国计算机学会优秀博士论文奖和吴文俊人工智能优秀青年奖，入选爱思唯尔2022中国高被引学者。

Seminar 2

✦

✦

✦

✦

✦

Topic: Human-in-the-loop Learning and Information Diffusion

Time & Date: 10:30-11:30 AM, December 18 (Monday)

Venue: E205, Administration Building East Wing

Speaker: Prof. Lingjie DUAN, Singapore University of Technology and Design

Host: Prof. Jianwei HUANG, The Chinese University of Hong Kong, Shenzhen 

Language: English

Abstract：

This talk introduces our recent studies on dynamic optimization and mechanism design for distributed agents to learn and share time-varying information in popular network scenarios. We first consider human-in-the-loop learning for congestion or routing games, where selfish agents make myopic routing decisions not only to jam each other but also miss exploration of non-shortest paths’ time-varying traffic conditions. We show that the existing platforms (e.g., Waze and Google Maps) to aggregate and publicize traffic information shared by agents just encourages their myopic routing, and may lead to huge efficiency loss. To incentivize selfish agents to reach the best exploitation-exploration trade-off, we propose a new selected information disclosure (SID) mechanism: we only reveal the latest traffic information to agents when they intend to over-explore the stochastic paths, while hiding such information when they want to under-explore. We prove that our mechanism at most doubles the efficiency loss of the social optimum.

We next consider human-in-the-loop information diffusion for viral marketing, where major social platforms (e.g., Facebook Marketplace and Pinduoduo) repeatedly invite different agents as seeds to share fresh promotion information with their friends. As such, we optimize a multi-stage seeding process of viral marketing in social networks, and adopt the recent notions of the peak and the average age of information (AoI) to measure the timeliness of promotion information received by all agents. Our problems are different from the literature on information diffusion in social networks, which limits to one-time seeding and overlooks information replacement over time. Given NP-hard nature of the problems, we present polynomial-time algorithms with good approximation guarantees and verify the theoretical results by experiments on a real social network.

Bios：

Lingjie DUAN is an Associate Professor and Associate Head of the Engineering Systems and Design Pillar at Singapore University of Technology and Design (SUTD), which was established in collaboration with MIT. He received Ph.D. degree from The Chinese University of Hong Kong in 2012, and Bachelor of Electrical Engineering from Harbin Institute of Technology in 2008. During 2011, he was a visiting scholar University of California at Berkeley. He has been actively working and contributing to the interdisciplinary research field “Network Economics”, by combining computer networking with economics and algorithmic game theory. He received 2016 SUTD Excellence in Research Award, and in 2015 he received the 10th IEEE ComSoc Asia-Pacific Outstanding Young Researcher Award. He is an Associate Editor of IEEE/ACM Transactions on Networking (TON) and IEEE Transactions on Mobile Computing (TMC), and was an Editor of both IEEE Transactions on Wireless Communications and IEEE Communications Surveys and Tutorials. He was also a Guest Editor of IEEE Journal on Selected Areas in Communications, IEEE Wireless Communications Magazine, and IEEE Transactions on Cognitive Communications and Networking. He also served as a General Chair of WiOpt 2023 Conference, Co-Chair of Cognitive Radio and Networks Symposium of IEEE ICC 2019, and Co-Chair of Future Trends and Emerging Technologies of VTC 2017.

Seminar 3

✦

✦

✦

✦

✦

Topic: Malliavin-Stein's method, Poincare's inequalities, with applications in the spatial average of SPDEs

Time & Date: 10:30-11:30 AM, December 19 (Tuesday)

Venue: Room 207, Cheng Dao Building

Speaker: Dr. Panqiu XIA, Auburn University

Host: Prof. Cheuk Yin LEE, The Chinese University of Hong Kong, Shenzhen 

Language：English

Abstract：

In this talk, I will provide a brief overview of the central limit theorem using the Malliavin-Stein approach, and an incorporation of Poincare's inequalities. Additionally, I will show an application of the Malliavin-Stein method in studying the spatial average of the solution to the parabolic/hyperbolic Anderson model. The content of this talk is based on several research projects conducted in collaboration with Raluca Balan, Jingyu Huang, David Nualart, Xiong Wang, Wangjun Yuan, and Guangqu Zheng.

Bios:

Dr. Panqiu XIA is currently a postdoc fellow in the Department of mathematics and Statistics at Auburn University, United States, under the guidance of Prof. Le CHEN. He started his current position in Fall 2022. Previously, he was a postdoc at University of Copenhagen, Denmark from 2020 to 2022 and his mentor was Prof. Carsten Wiuf. Dr. XIA obtained his Ph.D. degree in 2020 from University of Kansas under joint supervision by Prof. Yaozhong HU and Prof. David Nualart. He received his B.Sc and M.Sc degrees from Wuhan University in 2011 and Macau University in 2014, respectively.

Seminar 4

✦

✦

✦

✦

✦

Topic: Critical Issues in the Fabrication of Color Conversion Layers for Micro-LED Display via Electrohydrodynamic Inject Printing

Time & Date: 3:40-4:40 PM, December 19 (Tuesday)

Venue: Room 207, Cheng Dao Building

Speaker: Prof. Yue LIN, Xiamen University

Host: Prof. Wengang BI, The Chinese University of Hong Kong, Shenzhen 

Language：English

Abstract：

In the micro-LED display technology, the color conversion layer (CCL) is key to the full color rendering. So far, the electrohydrodynamic inject printing (EHD-IP) is proven to be of the highest material utilization with the highest spatial resolution. Still, the EHD-IP faces three main obstacles before its fully commercialization: a) The red pixel deficiency, including low efficiency and short life-span; b) Lack of homogeneity of the CCL pixels, such as the well-known “coffee-ring” effect; c) The blue light leakage, accompanied by the low light conversion efficiency. In this talk, in order to give some potential solutions, I will introduce new techniques that have been developed in our research group recently: The highly efficient red-emitting nano-complex; The dual non-polar solvent QD colloidal ink for EHD-IP; Three-dimensional micro-pillar arrays fabricated by EHD-IP. With these three solutions, we have the potential capability to fabricate commercially competent CCL arrays for micro-LED displays. Challenges remain as how to realize reliable mass production in EHD-IP by employing multiple nozzles without significantly reducing the production yield.

Bios：

Dr. Yue LIN, associate professor in the Department of Electronic Science, Xiamen University. His research scope includes a) Micro-LED display technology; b) Electrohydrodynamic inkjet printing for quantum dot color conversion layers; c) GaN and AlGaN based semiconductor material and devices and the optoelectronic properties; d) All-inorganic and hybrid perovskite quantum dots and the optoelectronic properties. As first and/or corresponding author, he published more than 20 papers on journals such as Adv. Mat., Adv. Opt. Mat., Small, Appl. Phys. Lett., Photonics Res., etc. He hosted fundings like NSF of China, NSF and cooperation of Fujian Province, etc. 

Seminar 5

✦

✦

✦

✦

✦

Topic: GaN-based Micro-LEDs for Advanced Display Application

Time & Date: 10:00-11:00 AM, December 21 (Thursday)

Venue: Tencent Meeting

Tencent Meeting ID：863 964 752

Password：123456

Speaker: Prof. Tao TAO, Nanjing University

Host: Prof. Wengang BI, The Chinese University of Hong Kong, Shenzhen

Language：Chinese

Abstract：

GaN based Ⅲ-Nitride materials have outstanding characters including high electric breakdown field, high-electron saturation velocity, good chemical stability, large thermal conductivity, high mobility and wide adjustable energy bandgap, thus have attracted significant interests in illumination and display society. GaN-based Micro-LEDs have become a research hotspot as a novel display technology due to its numerous unique advantages of high brightness, color saturation, fast response times, and long lifespan. Compared with traditional liquid crystal displays (LCDs) and organic light-emitting diodes (OLEDs), Micro-LEDs have been considered as one of the most ideal choices for a variety of novel display applications in virtual reality (VR), augmented reality (AR), optical communication, biomedical probes and etc. Of course, new applications require new developments, focusing on the following aspects: (1) high efficiency; (2) high luminance; (3) high resolution, which depend on high quality material, well designed quantum structures and advanced integration technique. In this report, recent progress in related area will be demonstrated and discussed.

Bios：

Prof. Tao TAO is from School of Electronic Science and Engineering, Nanjing University, Nanjing, China, focusing on GaN-based Micro-LEDs. Prof. TAO has published more than 70 peer reviewed papers, including Nature Nanotechnology, Advanced Functional Materials, IEEE Electron Device Letters etc.

【END】

点击以下链接，进入理工时刻：