DET Talks
What is it?
DET talks are an initiative of the DET Department to facilitate scientific interaction, bringing people together to attend technical talks by either external or internal experts. PhD students of the EECE doctoral program with registered attendance to a DET talk (google form registration + signature on attendance sheet) receive one hour of "hard skills" counting towards satisfying the EECE requirements. Plus, a coffee break is offered at the end of the talk.
EECE PhD student registration link (same for all talks, be sure to insert the correct talk date): here
Upcoming Talks:
Next one is going to be two short talks as follows:
Title: Large-scale Deep Reinforcement Learning with Serverless Computing
Speaker: Hao Wang, Stevens Institute of Technology, USA
Abstract:
Deep reinforcement learning (DRL) algorithms are widely applicable in many different areas, such as scientific simulations, robotics, autonomous driving, and large language model (LLM) development. However, DRL training is computationally expensive, requiring numerous trial-and-errors and consuming substantial computing resources and time. From an algorithmic perspective, the stochastic nature of environment dynamics can cause some actors to complete episodes sooner, resulting in idle periods while waiting for other actors to finish. From a systems perspective, actors remain idle during the policy update by the learner, significantly wasting computing resources and amplifying training costs. In this talk, I will introduce our recent studies published by AAAI’24 and accepted by SC’24. The SC’24 paper has received the Best Student Paper Nomination. First, I will address the fundamental challenges in large-scale distributed DRL training. I will then delve into our recent studies on distributed DRL training using serverless computing. Serverless computing, also known as Function-as-a-Service (FaaS), is a cloud computing model that employs lightweight containers as execution units. The instant execution and auto-scaling capabilities of serverless computing naturally meet the highly dynamic resource demands of DRL training. Finally, I will discuss our ongoing research and future outlook for serverless computing architectures and LLM inference.
Speaker Bio:
Dr. Hao Wang will join the Electrical and Computer Engineering Department at Stevens Institute of Technology this fall as an Assistant Professor. Previously, he served as an Assistant Professor in the Department of Computer Science and Engineering at Louisiana State University, beginning in Spring 2021. Dr. Wang holds a Bachelor’s and a Master’s degree from Shanghai Jiao Tong University, China, and a Ph.D. from the University of Toronto, Canada, advised by Prof. Baochun Li. His research focuses on integrating artificial intelligence with novel system architectures by designing robust AI algorithms and developing intelligent, high-performance computing systems. His work has been supported by six NSF grants (a total amount of $3 million) and industrial companies (e.g., IBM and AWS). He received the NSF CRII Award in 2022. His studies have been featured in prestigious conferences and journals, including ACM ASPLOS, ACM/IEEE SC, VLDB, IEEE INFOCOM, ACM MM, NeurIPS, ICLR, and AAAI.
--- followed by:
Title: Advances in NeRFs, Road Estimation, and Mapping for Automated Driving Systems
Speaker: Junsheng Fu, Zenseact, Sweden
Abstract:
This presentation explores recent advancements in Neural Radiance Fields (NeRFs), road estimation, and mapping technologies for Automated Driving Systems. We introduce NeuRAD, a novel view synthesis method tailored for dynamic autonomous driving data, showcasing the effectiveness of NeRFs in closed-loop simulations. Then, we discuss road estimation techniques, including model-based perceived road estimation, map-based road estimation, and online mapping, highlighting the challenges associated with each approach. Furthermore, we demonstrate the creation of lane-level maps utilizing Bayesian simultaneous localization and multi-lane tracking methods. Finally, we will present the Zenseact Open Dataset for research and development in autonomous driving applications.
Speaker Bio:
Junsheng Fu is a Technical Expert in Localization and Road Estimation at Zenseact, where he collaborates closely with teams focused on Localization, Road Estimation, and Mapping. He leads research projects, supervising industrial PhD students, PostDoc, and Master theses. Previously, Junsheng worked as a Computer Vision Researcher at Nokia Research in Finland. He holds a PhD in Computer Vision from Tampere University in Finland.About Zenseact: It is a Volvo Cars-owned self-driving software company. They are developing advanced safety software for AD and ADAS. Their technology encompasses every aspect from computer vision, sensor fusion, planning, and control, using a combination of rule-based and deep learning approaches. https://zenseact.com/
When & where: 7 October 2024, 3:00 PM, sala Luigi Ciminiera, DAUIN department (5th floor, entrance from C.so Castelfidardo 34D)
Flyer: DET Talk 07102024 
(273.41 kB)
----------------------------
Past Talks:
Title: Unlock the Potentials of Large-Element-Spacing Antenna Arrays: Grating-Lobe Suppression, High Gain, Low Sidelobe, and Beam Scanning
Speaker: Prof. Yuehe Ge, College of Physics and Information Engineering, Fuzhou University, China
Abstract:
Antenna arrays with large-element-spacing (LES, >1) offer advantages such as lower cost and less structural complexity. However, they typically suffer from high-level grating lobes, limiting their applications. Therefore, grating-lobe suppression technology in low-profile, high-gain planar antenna arrays has long been an interesting focus for antenna researchers.
In this talk, I will introduce a novel method to eliminate or reduce grating lobes in sparse, thinned, and uniformly large-element-spacing (LES) antenna arrays. Additionally, The radiation performance of the LES array, focusing on achieving high gain, high aperture efficiency, low sidelobes, and effective beam scanning, has been investigated. The results are promising and will be presented in detail. While our investigation is based on uniformly spaced LES antenna arrays, the conclusions drawn can also be applied to non-uniform, sparse, or thinned arrays.
Speaker Bio:
Yuehe Ge is a Professor in the College of Physics and Information Engineering at Fuzhou University, China. He received the Ph.D. degree from Macquarie University, Australia, in 2003. From 1991 to 1999, he was an Antenna Engineer at Nanjing Marine Radar Institute, China. From 2002 to 2011, he was a Research Fellow in the Department of Electronic Engineering, Macquarie University, Australia. In June 2011, he joined Huaqiao University, China, and became a Full Professor. Since July 2020, he has been a Professor at Fuzhou University, China. His research interests include antenna theory and designs for radar and communication applications, computational electromagnetics and optimization methods, metamaterials and metasurfaces as well as their applications. He has authored and co-authored over 200 journal and conference publications.
Professor Ge received several prestigious prizes from China State Shipbuilding Corporation and China Ship Research & Development Academy in 1995 and 1996. He received 2000 IEEE MTT-S Graduate Fellowship Awards. He is the co-winner of 2004 Macquarie University Innovation Awards-Invention Disclosure Award. He has served as a Guest Editor for the IEEE Transactions on Microwave Theory and Techniques and a technical reviewer for over 10 international journals and conferences. He was the General Co-Chair of the 2020 Cross-Strait Radio Science & Wireless Technology Conference (CSRSWTC2020) and the TPC Co-Chair of APCAP2020, APCAP2022, and APCAP2023.
When & where: 12 July 2024, 2:00 PM, Meeting Room DIMEAS P3
Flyer: DET Talk 07102024 (273.41 kB)
----------------------------
Title: Implantable Electronics for a High-Fidelity Artificial Retina
Speaker: Dr. Dante Gabriel Muratore, TUDelft (The Netherlands)
Abstract: Electronic interfaces to the retina represent an exciting opportunity to restore or even enhance vision. Although proof of principle devices have been demonstrated, they provide limited visual function. This is because they only provide coarse control over the targeted neural circuitry and fail to respect its cellular and cell-type specificity. To achieve better results, future devices should be able to control a large population of neurons with cellular resolution. In this talk, I will present the design of a wireless bi-directional neural interface and discuss on the circuit and system challenges associated with the design of its implantable electronics.
Speaker Bio: Dante G. received a B.Sc. and an M.Sc. degree in Electrical Engineering from Politecnico of Turin, Italy in 2012 and 2013, respectively. He received a Ph.D. degree in Microelectronics from the University of Pavia, Italy in 2017 in the Integrated Microsystems Lab. From 2015 to 2016, he was a Visiting Scholar at Microsystems Technology labs at the Massachusetts Institute of Technology, USA. From 2016 to 2020, he was a Postdoctoral Fellow at Stanford University, USA. He is the recipient of the Wu Tsai Neurosciences Institute Interdisciplinary Scholar Award. Since 2020, he is an assistant professor in the Bioelectronics Section at Delft University of Technology, Netherlands, where he leads the Smart Brain Interfaces group. His group investigates hardware and system solutions for high-bandwidth brain-machine interfaces that can interact with the nervous system at natural resolution. They contribute solutions for massively parallel bidirectional interfaces, on-chip neural signal processing, and wireless power and data transfer.
When & where: 18 June 2024, 10:00 AM, sala Luigi Ciminiera, DAUIN department (5th floor, entrance from C.so Castelfidardo 34D)
Flyer: DET talk 18062024 (262.73 kB)
----------------------------
Title: Dynamic Distributed Computing for Autonomous Vehicles
Speaker: Prof. Marco Levorato, University of California, Irvine, USA
Abstract: Neural networks are becoming a central component of a broad range of applications. However, the complexity of the tasks, the diversity of operating contexts, as well as channel and computing resource scarcity challenge the effective deployment of neural models in many relevant scenarios. In this talk, I will provide an overview of the techniques and frameworks that my research group developed to allow flexible, efficient and resilient distributed neural computing for robotic perception and autonomous navigation. Our approaches deeply integrate system and machine learning to obtain practical solutions deployable on real-world hardware platforms and applications.
Speaker Bio: Marco Levorato is a Professor in the Computer Science department at the University of California, Irvine. He completed the PhD in Electrical Engineering at the University of Padova, Italy, in 2009. Between 2010 and 2012, he was a postdoctoral researcher Jointly at Stanford and the University of Southern California. Prof. Levorato’s research interests are focused on distributed computing over unreliable wireless systems, especially for autonomous vehicles and robotic applications. In this area of research, he has more than 170 papers in IEEE and ACM venues. His work received the best paper award at IEEE GLOBECOM (2012). He received the UC Hellman Foundation Award in 2016, the Dean mid-career research award in 2019, and the UCI Innovator Award in 2024. His research is funded by the National Science Foundation, the Department of Defense, Intel and Cisco. In 2020-2021, he was the vice chair of the IEEE Technical Committee on Smart Grid Communications. He serves in the TPC of IEEE Infocom, IEEE Secon, IEEE Percom, IEEE ICDCS and ACM MobiHoc, is an editor of the IEEE/ACM Transactions on Networking, and was part of the organizing committee of several IEEE and ACM conferences.
When & where: 26 June 2024, 11:00 AM, sala Maxwell, DET department (5th floor)
Flyer: DET talk 26062024 (314.63 kB)
----------------------------