BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Department of Electrical and Computer Engineering (HKUECE) 電機與計算機工程系 - ECPv6.16.0//NONSGML v1.0//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
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:20230101T000000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTART;TZID=Asia/Hong_Kong:20240514T140000
DTEND;TZID=Asia/Hong_Kong:20240514T150000
DTSTAMP:20260512T175331
CREATED:20240503T092811Z
LAST-MODIFIED:20250114T063407Z
UID:18489-1715695200-1715698800@ece.hku.hk
SUMMARY:RPG Seminar – Complex-valued Transformer for Wireless Communications
DESCRIPTION:Meeting ID: 892 6022 4678\nPassword: n9QxYg \nAbstract\nIn recent years\, attention-based models\, particularly those employing the transformer structure\, have exhibited exceptional performance in tasks such as natural language processing\, computer vision\, and wireless communication\, among others. Notably\, most of these transformer frameworks rely on real-valued operations\, where inputs\, outputs\, and trainable parameters are real numbers\, even in tasks involving complex-valued domains.\nComplex-valued neural networks (CVNN) have emerged as a powerful approach for addressing problems associated with the complex-valued nature of input data. However\, the application of complex-valued transformers remains largely unexplored within the field of wireless communication\, where most task inputs\, such as received symbols and channel coefficients\, are intrinsically complex-valued. \nThis seminar aims to unveil the potential of complex-valued transformers for complex-valued domain tasks in wireless communications\, specifically focusing on channel estimation in single-input-single-output system and device activity detection in grant-free massive access scenario. To this end\, we propose tailored complex-valued transformer designs that incorporate complex-valued attention mechanisms for both tasks. In particular\, the proposed frameworks exploit the relationship between the real and imaginary parts of signals as implicit constraints\, while capturing temporal and spatial correlations of complex-valued input features. Moreover\, we introduced a novel complex-to-real layer to convert the complex-valued feature into the probabilistic representation for the activity detection task. \nNumerical results demonstrate the superiority of the proposed complex-valued transformers framework compared with other deep-learning based methods and optimization approaches. \nSpeaker\nMiss Leng Yang\nDepartment of Electrical and Electronic Engineering\,\nThe University of Hong Kong \nBiography of the speaker\nMiss Leng Yang received the B.Eng. degree in Electronic and Information Science and Technology from Fudan University in 2022. She is currently pursuing the MPhil degree with the Department of Electrical and Electronic Engineering\, The University of Hong Kong\, Hong Kong. \nOrganizer\nProf. Yik-Chung WU \nAll are welcome.
URL:https://ece.hku.hk/events/20240514-1/
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:20240514T140000
DTEND;TZID=Asia/Hong_Kong:20240514T150000
DTSTAMP:20260512T175331
CREATED:20240503T094824Z
LAST-MODIFIED:20250114T063447Z
UID:18490-1715695200-1715698800@ece.hku.hk
SUMMARY:RPG Seminar – Image Augmented Multimodal Autolabeller for 3D Object Detection
DESCRIPTION:Abstract\nRecently deep learning methods have gained groundbreaking success in many areas\, including autonomous driving and 3D object detection. Powerful neural networks are proposed and yield human-comparable ability after being trained from large datasets. Nonetheless\, the annotation procedure is time-consuming and tedious. To automate the annotation process\, we proposed two methods called MAP-Gen and MTrans\, respectively. Leveraging both image and point cloud modalities\, the two methods can effectively alleviate the sparsity problem of point clouds and hence generate high-quality pseudo labels. \nSpeaker\nMr. Liu Chang\nDepartment of Electrical and Electronic Engineering\,\nThe University of Hong Kong \nBiography of the speaker\nMr. Liu Chang received his B.Eng. degree in Computer Engineering from the University of Hong Kong. He is currently a Ph.D. student supervised by Dr. N. Wong and Prof. Edmund Y. Lam\, at the Department of Electrical and Electronic Engineering\, University of Hong Kong. His current research interests include 3D Vision\, Point Cloud\, Multi-modal Neural Networks\, and NLP. \nOrganizer\nProf. N. WONG
URL:https://ece.hku.hk/events/20240514-2/
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:20240514T140000
DTEND;TZID=Asia/Hong_Kong:20240514T150000
DTSTAMP:20260512T175331
CREATED:20240506T013049Z
LAST-MODIFIED:20250114T063519Z
UID:18492-1715695200-1715698800@ece.hku.hk
SUMMARY:RPG Seminar – Memristor Enabling Efficient Combinatorial Optimization with Quantum-inspired Parallel Annealing
DESCRIPTION:Abstract\nCombinatorial optimization problems are prevalent in various fields\, but obtaining exact solutions remains challenging due to the combinatorial explosion with increasing problem size. Special-purpose hardware such as Ising machines\, particularly memristor-based analog Ising machines\, have emerged as promising solutions. However\, existing simulate-annealing-based \nimplementations have not fully exploited the inherent parallelism and analog storage/processing features of memristor crossbar arrays. This work proposes a quantum-inspired parallel annealing method that enables full parallelism and improves solution quality\, resulting in significant speed and energy improvement when implemented in analog memristor crossbars. We experimentally solved tasks\, including unweighted and weighted Max-Cut and traveling salesman problem\, using our integrated memristor chip. The quantum inspired parallel annealing method implemented in memristor-based hardware has demonstrated significant improvements in time- and energy- efficiency compared to previously reported simulated annealing and Ising machine implemented on other technologies. This is because our approach effectively exploits the natural parallelism\, analog conductance states\, and all-to-all connection provided by memristor technology\, promising its potential for solving complex optimization problems with greater efficiency. \nSpeaker\nMr. Mingrui JIANG\nDepartment of Electrical and Electronic Engineering\,\nThe University of Hong Kong \nBiography of the speaker\nMr. Mingrui JIANG received the B.E. degree from School of Optical and Electronic Information\, Huazhong University of Science and Technology\, Wuhan\, China\, in 2020. He is currently pursuing the Ph.D. degree with the Department of Electrical and Electronic Engineering\, The University of Hong Kong\, Hong Kong SAR. His research interests include analog signal processing\, analog in-memory computing and neuromorphic computing based on emerging memory devices (e.g.\, memristors). \nOrganizer\nProf. Can LI \nAll are welcome.
URL:https://ece.hku.hk/events/20240514-3/
LOCATION:Online via Zoom
CATEGORIES:Seminar
ATTACH;FMTTYPE=image/jpeg:https://ece.hku.hk/wp-content/uploads/2024/11/rpg-seminar.jpg
END:VEVENT
END:VCALENDAR