UCLA VAST Lab的封面图片
UCLA VAST Lab

UCLA VAST Lab

研究服务

VLSI Architecture, Synthesis and Technology

关于我们

The VAST lab at UCLA investigates cutting-edge research topics at the intersection of VLSI technologies, design automation, architecture and compiler optimization at multiple scales, from micro-architecture building blocks, to heterogeneous compute nodes, and scalable data centers. Current focuses include architecture and design automation for emerging technologies, customizable domain-specific computing with applications to multiple domains, such as imaging processing, bioinformatics, data mining and machine learning. Please follow our page for posts on our latest news, and shared posts from the field. Our website gives additional info on the current group members, recent publications, and additional news items. Website: https://vast.cs.ucla.edu/ Youtube: https://www.youtube.com/channel/UC-MYfaosRizQaE7QggzfmOg

网站
https://vast.cs.ucla.edu/
所属行业
研究服务
规模
11-50 人
总部
Los Angeles
类型
教育机构

地点

UCLA VAST Lab员工

动态

  • 查看UCLA VAST Lab的组织主页

    521 位关注者

    Congratulations to Prof. Cong and his co-authors Congratulations to Prof. Jason Cong for his paper titled "Optimizing FPGA-based Accelerator Design for Deep Convolutional Neural Networks," published at the 23rd International Symposium on Field-Programmable Gate Arrays (FPGA'15) and co-authored with Chen Zhang, Peng Li, Guangyu?Sun, Yijin Guan, and Bingjun Xiao,?being inducted into the FPGA and Reconfigurable Computing Hall of Fame by the ACM SIGDA Technical Committee on FPGAs and Reconfigurable Computing.?This paper presented the first FPGA acceleration of deep neural networks, and has over 2,500 citations as of the end of 2024. This is the fifth paper co-authored by Prof. Cong to be inducted into the FPGA and Reconfigurable Computing Hall of Fame. For the original paper and the endorsement, please visit?https://lnkd.in/gReazyFw?

    • 该图片无替代文字
    • 该图片无替代文字
  • UCLA VAST Lab转发了

    Thrill to share that our work, “HMT: Hierarchical Memory Transformer for Efficient Long Context Language Processing”, has been accepted by #NAACL 2025 Main Conference! Highlights: - We proposed HMT, a novel plug-and-play framework for language models to enable and enhance their long-context processing ability. - HMT can achieve comparable or even better generation quality with 2~57x fewer parameters and 2.5x~116x lower memory requirements than large models with long context. - We organize information hierarchically, facilitating efficient memory retrieval based on contexts of ongoing conversations. - HMT consistently improves generation quality for various model architectures, including transformers and SSMs. This work further provides LLM accessibility to resource-constrained applications and represents another step forward to life-long language tasks. Huge thanks to the authors Yingqi Cao, Zongyue Qin, Neha Prakriya, Yizhou Sun, and Jason Cong!?UCLA VAST Lab Preprint:?https://lnkd.in/gCxNKxPk Code:?https://lnkd.in/gTEju-XP Looking forward to having discussions about the future of long-context language models at NAACL 2025! #NAACL2025 #LLM #ACL #NLP

  • 查看UCLA VAST Lab的组织主页

    521 位关注者

    Congratulations to Karl Marrett for winning the Best Student Paper Award at Brain Informatics 2024 Congratulations to Computer Science PhD graduate Karl Marrett (supervised by Prof. Jason Cong) for winning the?Best Student Paper Award titled "Gossamer: Scaling Image Processing and Reconstruction to Whole Brains"?at?Brain Informatics '24, the principal conference at the intersection of AI and brain imaging.?This work is based on interdisciplinary research between UCLA Computer Science (led by Prof. Jason Cong) and Medicine (led by Professors William Yang and Hongwei Dong), and produced an efficient segmentation platform of long-range, fine-grained cells amid terabyte scale 3D images.?The team also published another paper titled "High Throughput Training Label Generation from Whole Brain Images", also presented by Karl. The International Conference on Brain Informatics (BI) series has established itself as the world's premier research conference on Brain Informatics, which is an emerging interdisciplinary and multidisciplinary research field that combines the efforts of Cognitive Science, Neuroscience, Machine Learning, Data Science, Artificial Intelligence (AI), and Information and Communication Technology (ICT) to explore the main problems that lie in the interplay between human brain studies and informatics research.

    • 该图片无替代文字
  • 查看UCLA VAST Lab的组织主页

    521 位关注者

    Congratulations to Prof. Jason Cong for receiving the ISPD Lifetime Achievement Award Prof. Cong has been selected to receive the 2025 ISPD Lifetime Achievement Award, to be presented at the International Symposium on Physical Design (ISPD) in Austin, TX, on March 16-19, 2025.?This award is given to individuals who have made outstanding contributions to the field of physical design automation that span several decades. The purpose of the award is to recognize their lifetime contributions to research, education, and professional service. Each year, a special session of the symposium is reserved for the tribute to the award recipient.?For the list of previous winners, please see: https://lnkd.in/d-W93udC

相似主页