SUNMONTUEWEDTHUFRISAT
Events for March 02, 2022
-
ECE Seminar: Full Stack Deep Learning at the Edge
Wed, Mar 02, 2022 @ 10:00 AM - 11:00 AM
Ming Hsieh Department of Electrical and Computer Engineering
Conferences, Lectures, & Seminars
Speaker: Dr. Amir Gholami, Research Scientist, RiseLab and BAIR at UC Berkeley
Talk Title: Full Stack Deep Learning at the Edge
Abstract: An important next milestone in machine learning is to bring intelligence to the edge without relying on the computational power of the cloud. This could lead to more reliable, lower latency, and privacy preserving AI for a wide range of applications. However, state-of-the-art NN models require prohibitive amounts of compute, memory, and energy resources which is often not available at the edge. Addressing these challenges without compromising on accuracy, requires a multi-faceted approach, including hardware-aware model compression and accelerator co-design.
In this talk, I will first discuss a novel hardware-aware method for neural network quantization and pruning that achieves optimal trade-off between accuracy, latency, and model size. In particular, I will discuss a new Hessian Aware Quantization (HAWQ) method that relies on second-order information to perform low precision quantization of the model with minimal generalization loss. I will present extensive testing of the method on different learning tasks including various models for image classification, object detection, natural language processing, and speech recognition showing that HAWQ exceeds previous baselines. I will then present a recent extension of this method which allows integer-only inference for the end-to-end computations, enabling efficient deployment on fixed-point hardware. Finally, I will discuss a full-stack hardware-aware neural network architecture and accelerator design, which enables adapting the model architecture and the accelerator parameters to achieve optimal performance.
Related paper:
ICML'21: HAWQ-V3: Dyadic Neural Network Quantization
ICML'21: I-BERT: Integer-only BERT Quantization
Biography: Amir Gholami is a research scientist in RiseLab and BAIR at UC Berkeley. He received his PhD from UT Austin, working on large scale 3D image segmentation, a research topic which received UT Austin's best doctoral dissertation award in 2018. He is a Melosh Medal finalist, the recipient of best student paper award in SC'17, Gold Medal in the ACM Student Research Competition, best student paper finalist in SC'14, as well as Amazon Machine Learning Research Award in 2020. He was also part of the Nvidia team that for the first time made low precision neural network training possible (FP16), enabling more than 10x increase in compute power through tensor cores. That technology has been widely adopted in GPUs today. Amir's current research focuses on efficient AI, AutoML, and scalable training of Neural Network models.
Host: Host: Dr. Massoud Pedram, pedram@usc.edu
Webcast: https://usc.zoom.us/j/95064180366?pwd=SVJ3VzZ3aGNRKzNLdmJQeGRhdzBUZz09Location: Hughes Aircraft Electrical Engineering Center (EEB) - 248
WebCast Link: https://usc.zoom.us/j/95064180366?pwd=SVJ3VzZ3aGNRKzNLdmJQeGRhdzBUZz09
Audiences: Everyone Is Invited
Contact: Mayumi Thrasher
-
Center of Autonomy and AI, Center for Cyber-Physical Systems and the Internet of Things, and Ming Hsieh Institute Seminar Series
Wed, Mar 02, 2022 @ 11:00 AM - 12:00 PM
Ming Hsieh Department of Electrical and Computer Engineering
Conferences, Lectures, & Seminars
Speaker: Dimos V. Dimarogonas, School of Electrical Engineering and Computer Science, KTH Royal Institute of Technology
Talk Title: Multi-robot Task Planning and Control Under Spatiotemporal Specifications
Series: Center for Cyber-Physical Systems and Internet of Things
Abstract: Multi-robot task planning and control under temporal logic specifications has been gaining increasing attention in recent years due to its applicability among others in autonomous systems, manufacturing systems, service robotics and intelligent transportation. Initial approaches considered qualitative logics, such as Linear Temporal Logic, whose automata representation facilitates the direct use of model checking tools for correct-by-design control synthesis. In many real world applications however, there is a need to quantify spatial and temporal constraints, e.g., in order to include deadlines and separation assurance bounds. This led to the use of quantitative logics, such as Metric Interval and Signal Temporal Logic, to impose such spatiotemporal constraints. However, the lack of automata representations for such specifications hinders the direct use of model checking tools. Motivated by this, the use of transient control methodologies that fulfil the aforementioned qualitative constraints becomes evident. In this talk, we review some of our recent results in applying transient control techniques, and in particular Model Predictive Control, Barrier Certificates based design and Prescribed Performance Control, to distributed multi-robot task planning under spatiotemporal specifications. The results are supported by relevant experimental validations.
Biography: Dimos V. Dimarogonas received the Diploma in Electrical and Computer Engineering in 2001 and the Ph.D. in Mechanical Engineering in 2007, both from National Technical University of Athens (NTUA), Greece. Between 2007 and 2010, he held postdoctoral positions at the KTH Royal Institute of Technology, Dept of Automatic Control and MIT, Laboratory for Information and Decision Systems (LIDS). He is currently Professor at the Division of Decision and Control Systems, School of Electrical Engineering and Computer Science, at KTH. His current research interests include multi-agent systems, hybrid systems and control, robot navigation and manipulation, human-robot-interaction and networked control. He serves in the Editorial Board of Automatica and the IEEE Transactions on Control of Network Systems and is a Senior Member of IEEE. He is a recipient of the ERC Starting Grant in 2014, the ERC Consolidator Grant in 2019, and the Knut och Alice Wallenberg Academy Fellowship in 2015.
Host: Pierluigi Nuzzo, nuzzo@usc.edu
Webcast: https://usc.zoom.us/webinar/register/WN_zyIBh_1gQLmKpMJG0GyLxwLocation: Online
WebCast Link: https://usc.zoom.us/webinar/register/WN_zyIBh_1gQLmKpMJG0GyLxw
Audiences: Everyone Is Invited
Contact: Talyia White
