August 12, 2021 — The 4th Annual International Conference on Neuromorphic Systems, or iconHeld virtually from July 27th to 29th, 175 participants from four continents gathered to attend live sessions, view pre-recorded presentations, and of new architectures, models, and applications. I talked with my colleagues about pushing the limits.
Hosted annually by the Oak Ridge National Laboratory of the Department of Energy since 2017, ICONS provides the best platform for generating and explaining insights into the evolution of neuromorphic systems. This category of computing resources uses hardware and software designed to emulate the complex data analysis and decision-making capabilities of the human brain.
Program Co-Chair Katie Schuman and other scientists specializing in this area aim to provide an alternative to traditional computing methods. This requires not only new physical systems, but also more sophisticated algorithms and user-friendly software frameworks. Throughout the event, speakers emphasized the importance of developing and distributing these resources to enable hands-on experimentation.
Schuman, a computer scientist at ORNL, said: “In recent years, the community has changed and the variety of algorithms for advanced applications has exploded.”
The advantage of this approach is astronomically when the feasible amount of computing power available for research purposes finally stops growing in a stable orbit, a long-awaited scenario known as the end of Moore’s Law. There is a possibility of becoming.
Participants in ICONS 2021 watched a full-length presentation and a short “lightning talk”. One of them introduced a neuromorphic car that runs on a cardboard truck in Schumann’s basement. By testing driving performance in this controlled environment, Schumann and her collaborators, including Tom Potok General Chair and Robert Patton in the ORNL Data and AI Systems section, used neural networks. I want to better understand how to control self-driving cars.
Other activities included deep learning tutorial, A series of six poster discussions, a session focused on programs funded by DOE and other institutions, and a second installment payment for the PhD Consortium.
Designed for PhD Students nearing graduation will have participants explain their research, receive feedback from new mentors and potential employers, and career paths to different groups of panelists in this part of the conference. I was able to ask about. ICONS participants, including several ex-students from last year’s consortium, submitted 29 treatises to the minutes of the conference.
This year’s event also included the latest results session. In this session, five speakers will explain the state-of-the-art neuromorphic computing achievements achieved last month or even last week, and give a glimpse of progress to other members of the community before the results are summarized. I let you. In treatises, posters or other formal documents.
Some of the keynote speakers have summarized the state of the biologically inspired algorithm from different angles. Emre Neftci The University of California, Irvine and the Eurich Research Center focus on the benefits of traditional deep learning algorithms. Julie Glorie The Center National de la Recherche Scientifique / Thales Institute in France has determined the type of device needed to efficiently incorporate the algorithm into the hardware.
finally, Julia Sandamilskaya of Intel emphasizes the value of algorithms beyond those that rely on deep learning, describes her company’s efforts to become a one-stop shop for algorithm development, simulation, and hardware implementation, and a future software framework. I previewed.
“Neromorphic algorithms and applications other than deep learning are very important aspects of this area and are often overshadowed,” Schumann said. “We are inspired by biology and can do a great job without having to devote ourselves to deep learning as the only training method.”
Participants are all over the world, so organizers had to navigate significantly different time zones. Despite best efforts to limit sessions to business hours, a huge number of locales are inevitably unorthodox, from a 10 pm talk in Australia to a 4 am poster presentation in Singapore. It led to the timing. Still, the accessibility of virtual formats has attracted a wider audience than previous events.
“In the last two years of the conference, geographical diversity has increased significantly,” Schumann said. “In the past, we had occasional attendees from Europe and Asia, but we can reach more people online.”
To continue to encourage these global connections, ICONS has the potential to become a hybrid conference with virtual and face-to-face components in the future. This event was supported by DOE’s Advanced Scientific Computing Research program.
UT-Battelle LLC manages the Oak Ridge National Laboratory, DOE’s Department of Science, the largest supporter of basic research in physical sciences in the United States. DOE’s Department of Science is working to address some of the most pressing challenges of our time. For more information, please visit: https://energy.gov/science
Source: Elizabeth Rosenthal, ORNL
Brain-inspired computing concepts take the lead at ICONS conferences
Source link Brain-inspired computing concepts take the lead at ICONS conferences