Events for the 3rd week of March
-
ECE Seminar: The Role of Machine Learning in Electronic Design Automation
Mon, Mar 14, 2022 @ 10:00 AM - 11:00 AM
Ming Hsieh Department of Electrical and Computer Engineering
Conferences, Lectures, & Seminars
Speaker: Vidya A. Chhabria, Ph.D. Candidate, Electrical and Computer Engineering Department, University of Minnesota
Talk Title: The Role of Machine Learning in Electronic Design Automation
Abstract: For several decades, advances in hardware, accelerated by Moore's law and enabled by electronic design automation (EDA) tools, have sustainably met the demands for high computation at low energy and cost. However, emerging applications demand computing power far beyond today's system capabilities. Rapid advances in high-performance computing address the problem by using accelerators for specialized tasks such as machine learning (ML), increasing design diversity and system complexity. With Moore's law running out of steam, EDA tools now play a crucial role in meeting these computational demands. EDA tools are challenged to build chips that not only compensate for slow down in scaling, but also provide high performance for both ML and non-ML applications, which use a variety of new architectural techniques and operate under stringent performance constraints. Conventional EDA tools involve computationally expensive analysis and optimizations and are suboptimal as they often tradeoff speed for accuracy. ML promises to address these challenges as it has found tremendous success in solving these problems in classification, detection, and design space exploration problems in several different applications.
In this talk, I will show how leveraging ML techniques can revolutionize EDA tools by addressing the existing challenges. In particular, the talk will focus on tools that aid designers in (i) delivering power inside the chip without significant losses to meet power demands and (ii) sending the heat outside the chip to avoid high temperatures. The first section of the talk will show how a fast ML inference brings down several hours of runtime to a few milliseconds on industry-scale designs for these tasks. The second section will demonstrate how ML enables high-quality solutions through rapid optimizations. A key challenge with the proposed ML-based methods is the limited availability of open-source data and benchmarks for training and evaluation. The third section will show how ML can generate synthetic training sets and benchmarks for evaluating novel EDA solutions to these tasks. I will conclude by presenting avenues for future research in ML and EDA.
Biography: Vidya A. Chhabria is a Ph.D. candidate in the Electrical and Computer Engineering department at the University of Minnesota. She received her B.E. in Electronics and Communication from M. S. Ramaiah Institute of Technology, India, in 2016, and her M.S. in Electrical Engineering from the University of Minnesota in 2018. Her research interests are in the areas of electronic design automation, IC design, and machine learning. She has interned at Qualcomm Technologies, Inc. in the summer of 2017 and NVIDIA Corporation during the summers of 2020 and 2021. She received the ICCAD Best Paper Award in 2021, the University of Minnesota Doctoral Dissertation Fellowship in 2021, Louise Dosdall Fellowship in 2020, and Cadence Women in Technology Scholarship in 2020.
Host: Dr. Pierluigi Nuzzo, nuzzo@usc.edu
Webcast: https://usc.zoom.us/j/91321182725?pwd=ZDl0Qzc0b0F3cVRlZE1ORE11VHdCQT09Location: Hughes Aircraft Electrical Engineering Center (EEB) - 248
WebCast Link: https://usc.zoom.us/j/91321182725?pwd=ZDl0Qzc0b0F3cVRlZE1ORE11VHdCQT09
Audiences: Everyone Is Invited
Contact: Mayumi Thrasher
-
ECE Seminar: Data efficient high-dimensional machine learning
Wed, Mar 16, 2022 @ 10:00 AM - 11:00 AM
Ming Hsieh Department of Electrical and Computer Engineering
Conferences, Lectures, & Seminars
Speaker: Dr. Kamyar Azizzadenesheli, Assistant Professor, Department of Computer Science, Purdue University
Talk Title: Data efficient high-dimensional machine learning
Abstract: Traditional deep neural networks are maps between finite dimension spaces, and hence, are not suitable for modeling phenomena such as those arising from the solution of partial differential equations (PDE). In the first part of the talk, I introduce a new deep learning paradigm, called neural operators, that learns operators which are maps between infinite dimension spaces. I show that neural operators are universal approximators of operators and demonstrate a series of empirical successes of neural operators in natural sciences.
In the second part, I talk about the intersection of control theory and reinforcement learning and establish data-efficient learning and decision-making methods for generic dynamical systems. I conclude the talk by presenting empirical successes of these principled methods.
Biography: Kamyar Azizzadenesheli is an assistant professor at Purdue University, department of computer science, since Fall 2020. Prior to his faculty position, he was at the California Institute of Technology (Caltech) as a Postdoctoral Scholar at the Department of Computing + Mathematical Sciences. Before his postdoctoral position, he was appointed as a special student researcher at Caltech, working with ML and Control researchers at the CMS department and the Center for Autonomous Systems and Technologies. He is also a former visiting student researcher at Caltech. Kamyar Azizzadenesheli is a former visiting student researcher at Stanford University, and researcher at Simons Institute, UC. Berkeley. In addition, he is a former guest researcher at INRIA France (SequeL team), as well as a visitor at Microsoft Research Lab, New England, and New York. He received his Ph.D. at the University of California, Irvine.
Host: Dr. Salman Avestimehr, avestime@usc.edu
Webcast: https://usc.zoom.us/j/93153496285?pwd=SmE3clJMSm9OVmVoNWdhMW1SVlk4QT09Location: Hughes Aircraft Electrical Engineering Center (EEB) - 248
WebCast Link: https://usc.zoom.us/j/93153496285?pwd=SmE3clJMSm9OVmVoNWdhMW1SVlk4QT09
Audiences: Everyone Is Invited
Contact: Mayumi Thrasher
-
ECE Seminar: Machine Learning for Precision Health: A Holistic Approach
Thu, Mar 17, 2022 @ 10:00 AM - 11:00 PM
Ming Hsieh Department of Electrical and Computer Engineering
Conferences, Lectures, & Seminars
Speaker: Dr. Ahmed Alaa, Postdoctoral Associate, Broad Institute of MIT & Harvard, MIT CSAIL
Talk Title: Machine Learning for Precision Health: A Holistic Approach
Abstract: Machine learning (ML) methods, combined with large-scale electronic health databases, could enable a personalized approach to healthcare by improving patient-specific diagnosis, prognostic predictions, and treatment decisions. If successful, this approach would be transformative for clinical research and practice. In this talk, I will describe a holistic approach to ML for precision health that comprises a three-step procedure: (1) data characterization and understanding, (2) model development and (3) model deployment. Next, I will demonstrate one instantiation of this approach in the context of developing ML models for predicting patient-level response to therapies using observational data. I will focus on a multi-task learning model that uses Gaussian processes to estimate the causal effects of a treatment on individual patients and discuss its application in various disease areas. Finally, I will discuss exciting avenues for future work, including ML methods for learning from unannotated clinical data, generating synthetic data and integrating clinical knowledge into data-driven modeling.
Biography: Dr. Ahmed Alaa is a postdoctoral associate at Massachusetts Institute of Technology (MIT) and the Broad Institute of MIT and Harvard University. Previously, he was a joint postdoctoral scholar at Cambridge University, Cambridge Center for AI in Medicine and the University of California, Los Angeles (UCLA). He obtained his Ph.D. in Electrical and Computer Engineering from UCLA, where he was also a recognized (visiting) Ph.D. student at Oxford University. His research focuses on developing machine learning (ML) methods that can leverage healthcare data to enable a patient-centric approach to medicine, whereby ML models can inform disease diagnosis, prognosis and treatment decisions based on the characteristics of individual patients. He is the recipient of the (school-wide) 2021 Edward K. Rice Outstanding Doctoral Student Award at UCLA.
Host: Dr. Ashutosh Nayyar, ashutosn@usc.edu
Webcast: https://usc.zoom.us/j/94383946134?pwd=U1N4emFRaDBnc0pTd2VXUHMwSkVidz09Location: Hughes Aircraft Electrical Engineering Center (EEB) - 248
WebCast Link: https://usc.zoom.us/j/94383946134?pwd=U1N4emFRaDBnc0pTd2VXUHMwSkVidz09
Audiences: Everyone Is Invited
Contact: Mayumi Thrasher
-
ECE-EP Seminar - Mehdi Kiani, Thursday, March 17 at 2pm in EEB 248
Thu, Mar 17, 2022 @ 02:00 PM - 03:00 PM
Ming Hsieh Department of Electrical and Computer Engineering
Conferences, Lectures, & Seminars
Speaker: Mehdi Kiani, Pennsylvania State University
Talk Title: Wireless Hybrid Electrical-Acoustic Systems for Body-Machine Interface
Abstract: We have already witnessed significant efforts towards the research and development of neurotechnologies to radically enhance our understanding of the extremely complex central and peripheral nervous systems (CNS and PNS) by modulating and imaging their activities. These technologies can eventually be utilized in establishing body-machine interfaces (BMIs) with the CNS and PNS to offer effective, minimally invasive, and long-term solutions for neurological disorders and chronic disabilities such as spinal cord and brain injuries, stroke, Parkinson's disease, epilepsy, rheumatoid arthritis, and diabetes, to name a few. Despite all the developments over the past decade, closed-loop BMIs with minimally invasive high-spatiotemporal-resolution recording and stimulation capabilities from the large-scale distributed CNS/PNS circuits is still one of the grand challenges of the neuroscience research in the 21st century. In this talk, I will present our recent efforts (and future work) towards the development of advanced minimally invasive BMIs for modulating and sensing neural and electrophysiological activities with high spatiotemporal resolution at large scale. These BMIs are enabled by innovative integrated circuits, ultrasound, and wireless power/data (with different modalities such as ultrasound and magnetoelectric) technologies. I will particularly present two projects that leverage ultrasound beam focusing and steering with electronic beamforming to enable wireless implantable technologies for high-resolution, large-scale brain neuromodulation and gastric electrical-wave mapping.
Biography: Dr. Kiani received his Ph.D. degree in Electrical and Computer Engineering from the Georgia Institute of Technology in 2014. He joined the faculty of the School of Electrical Engineering and Computer Science at the Pennsylvania State University in August 2014 where he is currently an Associate Professor. His research interests are in the multidisciplinary areas of analog, mixed-signal, and power-management integrated circuits; ultrasound; and wireless power/data transfer and energy harvesting for wireless implantable medical devices and neural interfaces. He was a recipient of the 2020 NSF CAREER Award. He is currently an Associate Editor of the IEEE Transactions on Biomedical Circuits and Systems and IEEE Transactions on Biomedical Engineering. He also serves as a Technical Program Committee member of the IEEE International Solid-State Circuits Conference (ISSCC) in the IMMD subcommittee.
Host: ECE-Electrophysics
More Information: Mehdi Kiani Flyer.pdf
Location: Hughes Aircraft Electrical Engineering Center (EEB) - 248
Audiences: Everyone Is Invited
Contact: Marilyn Poplawski
-
ECE-EP Seminar - Najme Ebrahimi, Friday, March 18th at 10am in EEB 248
Fri, Mar 18, 2022 @ 10:00 AM - 11:00 AM
Ming Hsieh Department of Electrical and Computer Engineering
Conferences, Lectures, & Seminars
Speaker: Najme Ebrahimi, University of Florida
Talk Title: Next Generation Intelligent and Secured Wireless World: From IoT and Sensors to Wideband and Multi-band Scalable Circuit and System
Abstract: The future intelligent and secured wireless world needs connectivity at any time anywhere and under extreme conditions with over one trillion sensors and Internet-of-Things (IoT) devices connected to the network. To this end, the autonomous, and yet connected, wireless world is envisioned to provide sensing and high-data-rate communications, accurate localization and ranging, and resiliency. The major challenges to attain these goals are latency and energy efficiency requirements, that are largely affected by interference, multi-path, and channel fading. To tackle these challenges, wideband high frequency scalable arrays are desired to provide high data-rate communications and directional beams for interference cancellation. Furthermore, wideband/multiband circuits and systems are needed for accurate localization in the presence of severe multipath and fading in ultra-dense environments in IoT networks.
In this talk, firstly, I will present novel techniques to overcome the challenges for future wideband/multiband scalable transceiver arrays, including power-efficient local oscillator distribution and phase shifting, image selection architecture, and novel compact antenna-IC integration. I will then discuss our ongoing work towards the wideband/multiband signal generation and modulation for 6G and beyond as well as heterogonous integration of different technologies and modules for extending the Moore's law. Secondly, I will present multi-band circuit generation for IoT and sensor nodes to be employed in dense wireless networks. More specifically, I will present the first bidirectional circuitry for IoT transponder that reciprocally generates harmonics and subharmonics, covering two communication frequency bands interchangeably, which makes it a premier tool for localization and sensing protocols. I will also discuss future directions on advanced multi-band reconfigurable architecture for wireless sensors and IoTs compatible with network and physical layer protocols for security, communications, and localization.
Biography: Najme Ebrahimi is an Assistant Professor of Electrical and Computer Engineering at the University of Florida. Her research focuses on Mm-Wave/THz Scalable Array for high data rate communications and sensing as well as the security and connectivity of the next generation of distributed Internet-of-Things (IoT) networks. She was a post-doctoral research fellow at the University of Michigan- Ann Arbor from 2017 to 2020 under the departmental fellowship and earned her Ph.D. from the University of California, San Diego in June 2017. She was selected as a Rising Star by MIT EECS Rising Star program in 2019 and by ISSCC Rising Star program of the IEEE Solid-State Circuits Society in 2020. She is a member of the Microwave and Mm-Wave Integrated Circuits committee (MTT-14) and serves in the IMS2022 Technical Paper Review Committee (TPRC). She is the recipient of the 2021 DARPA Young Faculty Award (YFA).
Host: ECE-Electrophysics
More Information: Najme Ebrahimi Flyer.pdf
Location: Hughes Aircraft Electrical Engineering Center (EEB) - 248
Audiences: Everyone Is Invited
Contact: Marilyn Poplawski