BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Department of Electrical and Computer Engineering (HKUECE) 電機與計算機工程系 - ECPv6.15.20//NONSGML v1.0//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-WR-CALNAME:Department of Electrical and Computer Engineering (HKUECE) 電機與計算機工程系
X-ORIGINAL-URL:https://ece.hku.hk
X-WR-CALDESC:Events for Department of Electrical and Computer Engineering (HKUECE) 電機與計算機工程系
REFRESH-INTERVAL;VALUE=DURATION:PT1H
X-Robots-Tag:noindex
X-PUBLISHED-TTL:PT1H
BEGIN:VTIMEZONE
TZID:Asia/Hong_Kong
BEGIN:STANDARD
TZOFFSETFROM:+0800
TZOFFSETTO:+0800
TZNAME:HKT
DTSTART:20240101T000000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTART;TZID=Asia/Hong_Kong:20251126T140000
DTEND;TZID=Asia/Hong_Kong:20251126T150000
DTSTAMP:20260511T143724
CREATED:20251117T074555Z
LAST-MODIFIED:20251117T074555Z
UID:113912-1764165600-1764169200@ece.hku.hk
SUMMARY:RPG Seminar – Dynamic Motion Modeling and Planning of Fabric Piece
DESCRIPTION:Zoom Link: https://hku.zoom.us/j/9706928305?omn=95929409545 \nAbstract\nUnlike rigid objects\, fabric pieces are difficult for robots to plan motion because they are deformable objects with infinite degrees of freedom\, and their states evolve during robot motion. Instead of using a detailed model\, we propose using an oriented bounding box to approximate the state of the fabric piece. The fabric piece motion is approximated by a Transformer-based neural network. A simple yet effective robot trajectory is designed based on the predicted future motion of the fabric piece. Experimental results on an industrial robot system with a fabric piece demonstrate that the fabric piece can avoid collisions with different obstacles and types of fabric. We then extend this approach to garment dynamic motion planning\, incorporating more complicated oriented bounding box modeling and trajectory design methods. \nSpeaker\nMr. Letian Li\nDepartment of Electrical and Electronic Engineering\nThe University of Hong Kong \nBiography of the Speaker\nLetian Li received the B. Eng. degree in detection\, guidance\, and control technology and the M. Eng. degree in instrumentation science and technology from the School of Instrumentation and Optoelectronic Engineering\, Beihang University\, Beijing\, China\, in 2019 and 2022\, respectively. He is currently pursuing the Ph.D. degree with JC STEM Lab of Robotics for Soft Materials\, Department of Electrical and Electronic Engineering\, Faculty of Engineering\, The University of Hong Kong\, Hong Kong SAR\, China. He is engaged in collaborative research with the Centre for Transformative Garment Production\, Hong Kong SAR\, China. His research interests include motion planning and learning. \nOrganiser\nProf. Kazuhiro Kosuge\nDepartment of Electrical and Electronic Engineering\, The University of Hong Kong \nAll are welcome.
URL:https://ece.hku.hk/events/20251126-2/
LOCATION:Room CB-603\, 6/F\, Chow Yei Ching Building\, The University of Hong Kong
CATEGORIES:Seminar
ATTACH;FMTTYPE=image/jpeg:https://ece.hku.hk/wp-content/uploads/2024/11/rpg-seminar.jpg
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Asia/Hong_Kong:20251126T140000
DTEND;TZID=Asia/Hong_Kong:20251126T150000
DTSTAMP:20260511T143724
CREATED:20251119T043408Z
LAST-MODIFIED:20251119T043408Z
UID:113999-1764165600-1764169200@ece.hku.hk
SUMMARY:RPG Seminar – Hardware-Adaptive and Superlinear-Capacity Memristor-based Associative Memory
DESCRIPTION:Zoom Link: https://hku.zoom.us/j/99913066038?pwd=qqqBn1ojbFqbJJ4Koun6hucopMT2rJ.1 \nAbstract\nBrain-inspired computing aims to mimic cognitive functions like associative memory\, the ability to recall complete patterns from partial cues. Memristor technology offers promising hardware for such neuromorphic systems due to its potential for efficient in-memory analog computing. Hopfield Neural Networks (HNNs) are a classic model for associative memory\, but implementations on conventional hardware suffer from efficiency bottlenecks\, while prior memristor-based HNNs faced challenges with vulnerability to hardware defects due to offline training\, limited storage capacity\, and difficulty processing analog patterns. Here we introduce and experimentally demonstrate on integrated memristor hardware a new hardware-adaptive learning algorithm for associative memories that significantly improves defect tolerance and capacity\, and naturally extends to scalable multilayer architectures capable of handling both binary and continuous patterns. Our approach achieves 3x effective capacity under 50% device faults compared to state-of-the-art methods. Furthermore\, its extension to multilayer architectures enables superlinear capacity scaling (∝  for binary patterns) and effective recalling of continuous patterns (∝  scaling)\, as compared to linear capacity scaling for previous HNNs. It also provides flexibility to adjust capacity by tuning hidden neurons for the same-sized patterns. By leveraging the massive parallelism of the hardware enabled by synchronous updates\, it reduces energy by 8.8× and latency by 99.7% for 64-dimensional patterns over asynchronous schemes\, with greater improvements at scale. This promises the development of more reliable memristor-based associative memory systems and enables new applications research due to the significantly improved capacity\, efficiency\, and flexibility. \nSpeaker\nMr. Chengping He\nDepartment of Electrical and Electronic Engineering\nThe University of Hong Kong \nBiography of the Speaker\nChengping He received his B.Eng. and M.S. degrees from the Department of Physics at Nanjing University\, China\, in 2019 and 2022\, respectively. He is currently pursuing a Ph.D. in the Department of Electrical and Electronic Engineering under the supervision of Professor Can Li. His research focuses on in-memory computing\, analog computing\, associative memory\, and software-hardware co-design. \nOrganiser\nProf. Can Li\nDepartment of Electrical and Electronic Engineering\, The University of Hong Kong \nAll are welcome.
URL:https://ece.hku.hk/events/20251126/
LOCATION:Online via Zoom
CATEGORIES:Seminar
ATTACH;FMTTYPE=image/jpeg:https://ece.hku.hk/wp-content/uploads/2024/11/rpg-seminar.jpg
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Asia/Hong_Kong:20251126T150000
DTEND;TZID=Asia/Hong_Kong:20251126T160000
DTSTAMP:20260511T143724
CREATED:20251117T073459Z
LAST-MODIFIED:20251117T092027Z
UID:113908-1764169200-1764172800@ece.hku.hk
SUMMARY:RPG Seminar – A Novel Fabric Alignment System for Sewing
DESCRIPTION:Zoom Link: https://hku.zoom.us/j/9706928305?omn=95929409545 \nAbstract\nAccurate fabric alignment is essential in garment manufacturing\, yet remains a challenging and labor-intensive task. This work presents a novel automated fabric alignment system that integrates a vision-guided robotic platform and a new Global Local Weighted Iterative Closest Point (GLW-ICP) algorithm. The system estimates the pose of wrinkle-free fabric panels—even under partial occlusion—by aligning global fabric edges and local sewing lines to CAD models. A roller-based end-effector then manipulates the fabric to achieve millimeter-level alignment accuracy. Unlike traditional ICP methods\, GLW-ICP introduces adaptive weighting and sparsity to enhance robustness against occlusion and unmatched points. Experiments with various fabric types\, including shirts and collars\, demonstrate consistent\, high-precision alignment. This system reduces operator dependency\, improves consistency\, and serves as a crucial step toward fully automated garment production workflows. \n  \nSpeaker\nMr. Wenbo Dong\nDepartment of Electrical and Electronic Engineering\nThe University of Hong Kong \nBiography of the Speaker\nWenbo Dong received his B.Sc. in Automation from Northeastern University\, China\, and M.Sc. degrees in Control Engineering from Harbin Institute of Technology and Mechanical Engineering from the University of California\, Riverside. He is currently pursuing a Ph.D. at the University of Hong Kong\, where he is affiliated with the JC STEM Lab of Robotics for Soft Materials. \nOrganiser\nProfessor Kazuhiro Kosuge\nDepartment of Electrical and Electronic Engineering\, The University of Hong Kong \nAll are welcome.
URL:https://ece.hku.hk/events/20251126-1/
LOCATION:Room CB-603\, 6/F\, Chow Yei Ching Building\, The University of Hong Kong
CATEGORIES:Seminar
ATTACH;FMTTYPE=image/jpeg:https://ece.hku.hk/wp-content/uploads/2024/11/rpg-seminar.jpg
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Asia/Hong_Kong:20251126T160000
DTEND;TZID=Asia/Hong_Kong:20251126T170000
DTSTAMP:20260511T143724
CREATED:20251120T033601Z
LAST-MODIFIED:20251120T033601Z
UID:114032-1764172800-1764176400@ece.hku.hk
SUMMARY:RPG Seminar – Parameter-sharing AI Model Caching\, Delivery\, and Inference at the Edge
DESCRIPTION:Zoom Link:  https://hku.zoom.us/j/94531714904 \nAbstract\nThe rapid proliferation of AI applications in the 6G era calls for efficient support of edge intelligence\, where models must be cached and executed at the network edge to deliver low-latency inference services. Unlike cloud data centers with abundant resources\, edge servers are constrained in both storage and computation\, creating new bottlenecks in AI service provisioning. Two critical yet underexplored challenges are storage efficiency in edge caching and model loading during edge inference\, both of which fundamentally determine the efficiency of delivering AI services at the network edge. \nThis talk will present recent advances on parameter-sharing AI model edge caching and inference. We first introduce TrimCaching\, a framework that leverages parameter sharing across models to improve storage efficiency in edge caching and significantly enhance model downloading performance. Building on this foundation\, we then discuss PartialLoading\, which reduces the dominant latency from repeatedly loading model parameters into GPU memory by strategically scheduling user requests to reuse shared parameters. Together\, these two works establish a unified perspective on exploiting parameter sharing to mitigate both edge caching and inference bottlenecks\, paving the way for scalable and efficient edge intelligence in next-generation networks.\n \nSpeaker\nMr. Guanqiao Qu\nDepartment of Electrical and Electronic Engineering\nThe University of Hong Kong \nBiography of the Speaker\nMr. Guanqiao Qu received his B.E. and M.E. degrees in Electronics and Information Engineering from Harbin Institute of Technology (HIT) in 2020 and 2022\, respectively. He is currently pursuing the Ph.D. degree in the Department of Electrical and Electronic Engineering at the University of Hong Kong (HKU). His research interests include edge intelligence\, wireless networking\, distributed learning\, and edge inference. \nOrganiser\nProf. Xianhao Chen\nDepartment of Electrical and Electronic Engineering\, The University of Hong Kong \nAll are welcome.
URL:https://ece.hku.hk/events/20251126-3/
LOCATION:Online via Zoom
CATEGORIES:Seminar
ATTACH;FMTTYPE=image/jpeg:https://ece.hku.hk/wp-content/uploads/2024/11/rpg-seminar.jpg
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Asia/Hong_Kong:20251126T170000
DTEND;TZID=Asia/Hong_Kong:20251126T180000
DTSTAMP:20260511T143724
CREATED:20251121T024713Z
LAST-MODIFIED:20251121T024713Z
UID:114067-1764176400-1764180000@ece.hku.hk
SUMMARY:RPG Seminar – Split Learning: Empowering AI on Resource-Constrained Edge Devices
DESCRIPTION:Zoom Link: https://hku.zoom.us/j/94531714904 \nAbstract\nThe next-generation mobile network aims to natively support distributed intelligence\, such as federated learning\, across massive wireless edge devices. Unfortunately\, in the era of large models\, the deployment of federated learning faces significant obstacles due to the limited resources on edge devices. In this talk\, I will briefly introduce split learning (SL) and elucidate how it overcomes resource limitations via device-server co-training\, which transforms next-generation edge AI. Then\, I will present our recent work on adaptive split federated learning (AdaptSFL) in resource-constrained edge networks. Specifically\, our work first provides a unified convergence analysis of split federated learning (SFL) to quantify the impact of model splitting and client-side model aggregation on the learning performance\, based on which the AdaptSFL framework is developed to adaptively control model splitting and client-side model aggregation to balance communication-computing latency and training convergence in SFL. Simulations results demonstrate the effectiveness of our approach in accelerating SFL under resource constraints. At last\, I will conclude the talk by discussing open problems and challenges in SL at the wireless edge.\n \nSpeaker\nMr. Zheng Lin\nDepartment of Electrical and Electronic Engineering\nThe University of Hong Kong \nBiography of the Speaker\nZheng Lin is currently pursuing the Ph.D. degree with the Department of Electrical and Electronic Engineering\, The University of Hong Kong\, Hong Kong\, China.  His research interests include wireless networking\, edge intelligence\, and distributed machine learning. \nOrganiser\nProf. Xianhao Chen\nDepartment of Electrical and Electronic Engineering\, The University of Hong Kong \nAll are welcome.
URL:https://ece.hku.hk/events/20251126-4/
LOCATION:Online via Zoom
CATEGORIES:Seminar
ATTACH;FMTTYPE=image/jpeg:https://ece.hku.hk/wp-content/uploads/2024/11/rpg-seminar.jpg
END:VEVENT
END:VCALENDAR