Mon, Feb 14, 2022 @ 10:00 AM - 11:00 AM
Ming Hsieh Department of Electrical and Computer Engineering
Conferences, Lectures, & Seminars
Speaker: Dr. Ramyad Hadidi, Machine Learning Researcher, SK hynix America
Talk Title: Dealing with Data Deluge in the Edge Systems
Abstract: Each day, a huge amount of data is generated. Edge systems--any computing agent but large-scale datacenter machines--not only fuel this data deluge but also play an increasingly vital role in processing them. Unlike conventional systems that are engineered with abundant resources, edge systems operate in real-world conditions facing several unknown design-space trade-offs with limited resources restricting their full-scale autonomy. This leads to isolated, time-consuming, and costly approaches to each challenge that result in ad-hoc edge systems but not the optimal one. To effectively maneuver the constraints and unique multi-dimensional design space of edge systems, my research develops novel machine learning techniques and exploits hardware-software synergy by setting roadmaps across the hardware-software stack for the next generation of edge systems.
In my talk, first, with an example of quadcopter drones, I show how my research is a pioneer that reveals the unique multi-dimensional design space of edge systems and suggests optimal points within this space depending on the use case. By formulating the fundamental drone subsystems and introducing our open-source customizable drone, I explain how computations impact this design space. As an example of optimized computations, by exploring implementations of simultaneous localization and mapping (SLAM) on various hardware platforms (CPU, GPU, FPGA, and ASIC), I demonstrate which implementation is more reasonable for drones. The second part of my talk emphasizes the necessity of modern machine learning techniques, such as those utilizing heavy neural networks, in comprehending complex raw data in edge systems and acting upon the outcomes. I show how my research empowers edge devices to break their individual resource constraints by distributing the computation on collaborating peer devices and proposes edge-aware neural networks by exploring hardware-software co-designs, algorithmic modifications, and system-level optimizations. In the end, I propose my plans for effectively handling data in exotic frontiers of edge systems with unique constraints to stimulate thought-provoking applications for our future.
Biography: Ramyad Hadidi is currently a machine learning researcher at SK hynix working at the intersection of hardware, software, and edge devices, focusing on the efficient execution of deep learning algorithms. Ramyad Hadidi received his Ph.D. in computer science from Georgia Institute of Technology in May 2021 under the supervision of Professor Hyesoon Kim with his thesis titled "Deploying Deep Neural Networks in Edge with Distribution." Ramyad's research interests include but are not limited to computer architecture, robotics, edge computing, and machine learning. Besides his dissertation research, Ramyad has contributed to research on processing-in-memory, GPU systems, and hardware accelerators for sparse problems, believing a balance between depth and breadth leads to genuine research problems.
Host: Dr. Peter Beerel, firstname.lastname@example.org
Audiences: Everyone Is Invited
Contact: Mayumi Thrasher