| 日時(Date) |
2024年8月28日 (木) / August 28th, 2024 4限 (13:30--15:00) / 3rd period (13:30--15:00) |
|---|---|
| 場所(Location) | エーアイ大講義室（L1）, AI Inc. Seminar Hall （L1）+ Online (HYBRID) |
| 司会(Chair) | Yirong Kan |
| 講演者(Presenter) | Jinjun Xiong (University at Buffalo) / Jinjun Xiong (University at Buffalo) |
| 題目(Title) | A Statistical Distribution-based Deep Neural Network Model – a new perspective on effective learning |
| 概要(Abstract) | The impressive results achieved by deep neural networks (DNNs) in various tasks, computer vision in particular, such as image recognition, object detection and image segmentation, have sparked the recent surging interests in artificial intelligence (AI) from both the industry and the academia alike. The wide adoption of DNN models in real-time applications has, however, brought up a need for more effective training of an easily parallelizable DNN model for low latency and high throughput. This is particularly challenging because of DNN's deep structures. To address this challenge, we observe that most of existing DNN models operate on deterministic numbers and process one single frame of image at a time, and may not fully utilize the temporal and contextual correlation typically present in multiple channels of the same image or adjacent frames from a video. Based on well-established statistical timing analysis foundations from the EDA domain, we propose a novel statistical distribution-based DNN model that extends existing DNN architectures but operates directly on correlated distributions rather than deterministic numbers. This new perspective of training DNN has resulted in surprising effects on achieving not only improved learning accuracy, but also reduced latency and increased high throughputs. Preliminary experimental results on various tasks, including 3D Cardiac Cine MRI segmentation, showed a great potential of this new type of statistical distribution-based DNN model, which warrants further investigation. |
| 講演言語(Language) | 英語 (English) |
| 講演者紹介(Introduction of Lecturer) | Dr. Jinjun Xiong is Empire Innovation Professor with the Department of Computer Science and Engineering at University at Buffalo (UB). He also serves as the Scientific Director for the $20 M National AI Institute for Exceptional Education ((website link)), and Director for the SUNY-UB Institute for Artificial Intelligence and Data Science ((website link)). Prior to that, he was a Senior Researcher and Program Director for AI and Hybrid Clouds Systems at the IBM Thomas J. Watson Research Center. He was the former co-founder and co-director for the IBM-Illinois Center for Cognitive Computing Systems Research (C3SR), the success of which in 5 years has led to the 10-year $200M expansion of the center to the IBM-Illinois Discovery Accelerator Institute. His research interests are on across-stack AI systems research, including AI applications, algorithms, tooling and computer architectures. Many of his research results have been adopted in IBM’s products and tools. He published more than 160 peer-reviewed papers in top AI conferences and systems conferences. His publication won 9 Best Paper Awards and 9 Nominations for Best Paper Awards. He also won top awards from various international competitions, including the championship award for the IEEE GraphChallenge on accelerating sparse neural networks in 2020, and the First Place Awards for the 2019 DAC Systems Design Contest on designing an object detection neural network for edge FPGA and GPU devices, respectively. |