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:20240429T100000
DTEND;TZID=Asia/Hong_Kong:20240429T110000
DTSTAMP:20260512T175356
CREATED:20240419T082617Z
LAST-MODIFIED:20250114T064421Z
UID:18341-1714384800-1714388400@ece.hku.hk
SUMMARY:RPG Seminar – Uncertainty Quantification
DESCRIPTION:Abstract:\nUncertainty quantification plays a crucial role in electromagnetic compatibility and inference (EMC/EMI) in the field. Traditionally\, methods such as the Monte Carlo method\, stochastic Galerkin method\, stochastic collocation\, and linear regression have been developed to tackle the challenges of uncertainty quantification (UQ) problems. However\, these methods often face the issue of curse of dimensionality. In this study\, we propose two different approaches to quantify the uncertainty in EMC/EMI for partial equivalent element circuits. The first method utilizes interval analysis to establish the bounds of the quantities of interest. To further capture the stochastic parameters\, we employ the physical-informed neural network to construct the polynomial chaos expansion. Consequently\, the coefficients of the polynomial bases can be obtained. To construct a network without relying on computational models\, we employ the Wasserstein generative adversarial network with a gradient penalty to estimate the stochastic characteristics. This approach allows us to effectively estimate the uncertain properties without explicitly relying on a computational model. \nSpeaker:\nMs. Yuan Ping\nDepartment of Electrical and Electronic Engineering\,\nThe University of Hong Kong \nBiography of the speaker:\nYuan Ping received the B.E. degree and M. S. degree from Xidian University in 2016 and 2019. She is currently pursuing the Ph.D. degree with the Department of Electrical and Electronic Engineering\, The University of Hong Kong\, Hong Kong. Her research interests Uncertainty Quantification\, Phase retrieval and computational electromagnetic. \nOrganizer:   Prof. Lawrence K. YEUNG \nAll are welcome.
URL:https://ece.hku.hk/events/20240429-3/
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:20240429T110000
DTEND;TZID=Asia/Hong_Kong:20240429T120000
DTSTAMP:20260512T175356
CREATED:20240418T011701Z
LAST-MODIFIED:20250114T064341Z
UID:18277-1714388400-1714392000@ece.hku.hk
SUMMARY:RPG Seminar – Enhancing Performances of InGaN-MQW Thin-film Microdisk Laser with Hybrid ODRs
DESCRIPTION:Abstract:\nThe conventional GaN microdisk laser provides poor overlap between the whispering gallery modes (WGM) and the multi-quantum well (MQW) gain region. The thin-film microdisk structure was proposed to overcome this shortcoming\, but the absorptive nature of the metallic bonding layer cum mirror compromises optical confinement. In this work\, a dielectric distributed Bragg reflector (DBR) is integrated with the metallic mirror to form an omni-directional reflector (ODR) that provides high optical reflectance across a wide range of incidence angle to promote optical confinement in the microdisk. Optical-pumped lasing with average lasing threshold power density of 46.5 W/cm2 and Q factors of 18200 (near threshold) is achieved\, representing a major advancement in GaN microdisk laser technology. \nSpeaker:\nMiss Zhongqi WANG\nDepartment of Electrical and Electronic Engineering\nThe University of Hong Kong \nBiography of the speaker:\nMiss Zhongqi WANG received her B.Eng. degree in Materials Science and Engineering at Tsinghua University and now pursuing the Ph.D. degree in the Department of Electrical and Electronic Engineering at the University of Hong Kong. She is now working on GaN based laser fabrication GaN-based microdisk fabrication and analysis of lasing characteristics. \nOrganizer:\nProf. A.H.W. CHOI
URL:https://ece.hku.hk/events/20240429-2/
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:20240429T140000
DTEND;TZID=Asia/Hong_Kong:20240429T150000
DTSTAMP:20260512T175356
CREATED:20240411T013043Z
LAST-MODIFIED:20250114T064304Z
UID:18251-1714399200-1714402800@ece.hku.hk
SUMMARY:RPG Seminar – Learning A Low-Rank Feature Representation: Achieving Better Trade-Off between Stability and Plasticity in Continual Learning
DESCRIPTION:Speaker:\nMr. Zhenrong LIU\nDepartment of Electrical and Electronic Engineering\,\nThe University of Hong Kong \nAbstract:\nDeep neural networks require the ability to continually learn and adapt to real-world conditions. This ability\, called “continual learning\,” is essential for AI systems to evolve. Among various representative continual learning methods\, null-space-projection-based algorithms have attracted considerable research interest due to their faster training speed and lower memory requirements. These algorithms optimize network parameters in the null space of past tasks’ feature representation matrices\, ensuring stability. However\, quantitatively analyzing the balance between stability and plasticity in null-space-projection-based methods poses significant challenges\, complicating efforts to refine and improve such approaches. \nIn this seminar\, we comprehensively examine null-space-projection-based continual learning methods and uncover two essential insights. Firstly\, to maintain stability\, the rank of the feature covariance increases with the number of continual learning tasks\, leading to a reduction in the dimension of the feature covariance’s null space. Secondly\, the dimension of the feature covariance’s null space significantly influences the plasticity of continual learning. Building on these insights\, we quantitatively demonstrate the stability-plasticity relationship inherent in null-space-projection-based continual learning methods. Then\, based on the stability-plasticity relationship\, we introduce a novel training algorithm named LRFR (Low-Rank Feature Representation) to enhance plasticity without compromising stability. Specifically\, we judiciously select only a subset of neurons in each layer of the network while training individual tasks to learn the past tasks’ feature representation matrix in low-rank. This increases the null space dimension when designing network parameters for subsequent tasks\, thereby enhancing the plasticity. Using CIFAR-100 and TinyImageNet as benchmark datasets for continual learning\, the proposed approach consistently outperforms state-of-the-art methods. \nBiography of the speaker:\nMr. Zhenrong LIU received the B.E. degree in automation from Northeastern University in 2018 and the M.E. degree in information and communication engineering from the Southern University of Science and Technology 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. \nOrganizer: Prof. Yik-Chung WU \nAll are welcome!
URL:https://ece.hku.hk/events/20240429-1/
LOCATION:Online via Zoom
CATEGORIES:Seminar
ATTACH;FMTTYPE=image/jpeg:https://ece.hku.hk/wp-content/uploads/2024/11/rpg-seminar.jpg
END:VEVENT
END:VCALENDAR