Machine Learning at the Speed of Light

Electrical & Computer

PITTSBURGH (Jan. 6, 2021) — As we enter the next chapter of the digital age, data traffic continues to grow exponentially. To further enhance artificial intelligence and machine learning, computers will need the ability to process vast amounts of data as quickly and as efficiently as possible. Conventional computing methods are not up to the task, but in looking for a solution, researchers have seen the light—literally. Light-based processors, called photonic processors, enable computers to complete complex calculations at incredible speeds. New research published this week in the journal Nature examines the potential of photonic processors for artificial intelligence applications. The results demonstrate for the first time that these devices can process information rapidly and in parallel, something that today’s electronic chips cannot do. “Neural networks ‘learn’ by taking in huge sets of data and recognizing patterns through a series of algorithms,” explained Nathan Youngblood, assistant professor of electrical and computer engineering at the University of Pittsburgh Swanson School of Engineering and co-lead author. “This new processor would allow it to run multiple calculations at the same time, using different optical wavelengths for each calculation. The challenge we wanted to address is integration: How can we do computations using light in a way that’s scalable and efficient?” The fast, efficient processing the researchers sought is ideal for applications like self-driving vehicles, which need to process the data they sense from multiple inputs as quickly as possible. Photonic processors can also support applications in cloud computing, medical imaging, and more. “Light-based processors for speeding up tasks in the field of machine learning enable complex mathematical tasks to be processed at high speeds and throughputs,” said senior co-author Wolfram Pernice at the University of Münster. “This is much faster than conventional chips which rely on electronic data transfer, such as graphic cards or specialised hardware like TPUs (Tensor Processing Unit).” The research was conducted by an international team of researchers, including Pitt, the University of Münster in Germany, the Universities of Oxford and Exeter in England, the École Polytechnique Fédérale (EPFL) in Lausanne, Switzerland, and the IBM Research Laboratory in Zurich. The researchers combined phase-change materials—the storage material used, for example, on DVDs—and photonic structures to store data in a nonvolatile manner without requiring a continual energy supply. This study is also the first to combine these optical memory cells with a chip-based frequency comb as a light source, which is what allowed them to calculate on 16 different wavelengths simultaneously. In the paper, the researchers used the technology to create a convolutional neural network that would recognize handwritten numbers. They found that the method granted never-before-seen data rates and computing densities. “The convolutional operation between input data and one or more filters – which can be a highlighting of edges in a photo, for example – can be transferred very well to our matrix architecture,” said Johannes Feldmann, graduate student at the University of Münster and lead author of the study. “Exploiting light for signal transference enables the processor to perform parallel data processing through wavelength multiplexing, which leads to a higher computing density and many matrix multiplications being carried out in just one timestep. In contrast to traditional electronics, which usually work in the low GHz range, optical modulation speeds can be achieved with speeds up to the 50 to 100 GHz range.” The paper, “Parallel convolution processing using an integrated photonic tensor core,” (DOI: 10.1038/s41586-020-03070-1) was published in Nature and coauthored by Johannes Feldmann, Nathan Youngblood, Maxim Karpov, Helge Gehring, Xuan Li, Maik Stappers, Manuel Le Gallo, Xin Fu, Anton Lukashchuk, Arslan Raja, Junqiu Liu, David Wright, Abu Sebastian, Tobias Kippenberg, Wolfram Pernice, and Harish Bhaskaran.
Maggie Pavlick

University of Pittsburgh Joins New DOE Cybersecurity Manufacturing Innovation Institute

Electrical & Computer, Industrial, MEMS, Nuclear

SAN ANTONIO, TX (November 19, 2020) ... The University of Texas at San Antonio (UTSA) today formally launched the Cybersecurity Manufacturing Innovation Institute (CyManII), a $111 million public-private partnership. Led by UTSA, the university will enter into a five-year cooperative agreement with the U.S. Department of Energy (DOE) to lead a consortium of 59 proposed member institutions in introducing a cybersecure energy-ROI that drives American manufacturers and supply chains to further adopt secure, energy-efficient approaches, ultimately securing and sustaining the nation’s leadership in global manufacturing competitiveness.U.S. manufacturers are one of the top targets for cyber criminals and nation-state adversaries, impacting the production of energy technologies such as electric vehicles, solar panels and wind turbines. Integration across the supply chain network and an increased use of automation applied in manufacturing processes can make industrial infrastructures vulnerable to cyber-attacks. To protect American manufacturing jobs and workers, CyManII will transform U.S. advanced manufacturing and make manufacturers more energy efficient, resilient and globally competitive against our nation’s adversaries.“The University of Pittsburgh is proud to be among the inaugural member institutions of this national effort to develop cyber security and energy research to benefit U.S. manufacturing expertise,” noted Rob A. Rutenbar,Senior Vice Chancellor for Research at Pitt. “Both our Swanson School of Engineering and School of Computing and Information at the forefront of innovations in advanced manufacturing, cyber infrastructure and security, sustainable energy, materials science and supply chain management. Our faculty are looking forward to participating in this groundbreaking institute.”“The exploitation of advanced materials and computing can provide us with a more holistic approach to secure the nation’s manufacturing infrastructure, from communication networks and assembly lines to intricate computer code and distribution systems,” added Daniel Cole, Associate Professor of Mechanical Engineering and Materials Science and co-director of the Swanson School’s Hacking for Defense program. “Just as our personal computers and cell phones are vulnerable to cyber-attacks, so too is our complex manufacturing industry. But thanks to this national effort through CyManII, we will not only be able to develop defenses but also create more sustainable and energy efficient technologies for industry.”“I am excited for the potential collaborations between our faculty and the innovations they will develop,” said David Vorp, Associate Dean for Research at the Swanson School. “We already have a healthy collaboration with faculty in the School of Computing and Information, and sustainability informs our research, academics, and operations. CyManII presents a new opportunity for us to engage in transformative, trans-disciplinary research.”As part of its national strategy, CyManII will focus on three high priority areas where collaborative research and development can help U.S. manufacturers: securing automation, securing the supply chain network, and building a national program for education and workforce development. “As U.S. manufacturers increasingly deploy automation tools in their daily work, those technologies must be embedded with powerful cybersecurity protections,” said Howard Grimes, CyManII Chief Executive Officer and UTSA Associate Vice President and Associate Vice Provost for Institutional Initiatives. “UTSA has assembled a team of best-in-class national laboratories, industry, nonprofit and academic organizations to cybersecure the U.S. manufacturing enterprise. Together, we will share the mission to protect the nation’s supply chain, preserve its critical infrastructure and boost its economy.”CyManII’s research objectives will focus on understanding the evolving cybersecurity threats to greater energy efficiency in manufacturing industries, developing new cybersecurity technologies and methods, and sharing information and knowledge with the broader community of U.S. manufacturers.CyManII aims to revolutionize cybersecurity in manufacturing by designing and building a secure manufacturing architecture that is pervasive, unobtrusive and enables energy efficiency. Grimes says this industry-driven approach is essential, allowing manufacturers of all sizes to invest in cybersecurity and achieve an energy ROI rather than continually spending money on cyber patches.These efforts will result in a suite of methods, standards and tools rooted in the concept that everything in the manufacturing supply chain has a unique authentic identity. These solutions will address the comprehensive landscape of complex vulnerabilities and be economically implemented in a wide array of machines and environments.“CyManII leverages the unique research capabilities of the Idaho, Oak Ridge and Sandia National Laboratories as well as critical expertise across our partner cyber manufacturing ecosystem,” said UTSA President Taylor Eighmy. “UTSA is proud and honored to partner with the DOE to advance cybersecurity in energy-efficient manufacturing for the nation.”CyManII has 59 proposed members including three Department of Energy National Laboratories (Idaho National Laboratory, Oak Ridge National Laboratory, and Sandia National Laboratories), four Manufacturing Innovation Institutes, 24 powerhouse universities, 18 industry leaders, and 10 nonprofits. This national network of members will drive impact across the nation and solve the biggest challenges facing cybersecurity in the U.S manufacturing industry.CyManII is funded by the Office of Energy Efficiency and Renewable Energy’s Advanced Manufacturing Office (AMO) and co-managed with the Office of Cybersecurity, Energy Security, and Emergency Response (CESER). ------ Learn more about the Cybersecurity Manufacturing Innovation Institute.
Author: EmilyGuajardo, CyManII Communications Manager

Tracking Monarch Butterfly Migration with the World’s Smallest Computer

Electrical & Computer

Each year, monarch butterflies make their way from the U.S. and Canada to central Mexico, where they'll spend the winter. Inhee Lee, assistant professor of electrical and computer engineering at the University of Pittsburgh's Swanson School of Engineering, partnered with researchers at the University of Michigan to create a tiny sensor to track and monitor the environmental conditions the butterflies encounter on the way. "We had to create a sensor small enough to be glued to the butterfly, which presented challenges for how to power it," said Lee. "We created a sensor that operates on very little power, has a small battery, and contains a very small solar panel to recharge the battery." The information collected by the sensor will help researchers understand the environmental conditions along the butterflies' path and inform where to focus conservation activities. Story originally appeared on the Michigan Engineering News Center from The University of Michigan. Reposted with permission. ### In Mexico, days before the COVID-19 shutdown, a team of engineers and biologists were riding on horseback into the heart of a popular overwintering site for monarch butterflies to conduct preliminary tests on their newest iteration of the Michigan Micro Mote (M3). The project, supported in part by National Geographic, hopes to aid wildlife conservation efforts by shedding light on butterfly migration and habitat conditions. The M3, created by David Blaauw, Kensall D. Wise Collegiate Professor of EECS, and several other University of Michigan researchers, is a fully energy-autonomous computing system that acts as a smart sensing system and can be configured for a wide variety of applications. For this project, the M3 will be glued to the back of individual monarch butterflies to track and monitor environmental conditions – specifically light and temperature and eventually air pressure – they encounter during migration. “This is our most complex M3 system,” says Blaauw. “We need to capture data about the light intensity that is accurate down to a few seconds, and we need to be able to transmit that captured data a long distance because we will not be able to physically retrieve the specimens.” Blaauw and Prof. Inhee Lee, an ECE alum who is now at the University of Pittsburgh, are responsible for the chip and system design. Prof. Hun-Seok Kim designs and trains the algorithms that analyze the captured data and reconstruct the migratory path of the specimen. André Green, a professor of Ecology and Evolutionary Biology at U-M, analyzes these paths to learn more about monarch biology and applies this knowledge to conservation efforts. Monarchs can travel as far as 3,000 miles during migration, spending the summer across the US and southern Canada to breed and the winter mostly in Mexico and along the coasts of California. The sensors have to be hardy enough to survive the long trip, as well as any inclement weather along the way, but light enough so they don’t disrupt the behavior or harm the butterflies. This iteration of the M3 is the lightest yet, weighing around 50 milligrams, which is tenfold lighter than the lightest tracking devices to date. As part of the team’s preliminary tests, they attached M3s to several butterflies and monitored their condition in a greenhouse. “All initial indications are that we’re not having strong negative effects on the butterflies,” says Green. “We found no significant difference in their metabolism whether they were carrying the sensor or not.” The conventional method to study monarch migration involves attaching a paper tag to an individual butterfly and recovering the specimen at known monarch destinations. “Using that technique, we can know only the starting point and ending point for the specimens we recover, which is a small percentage of the total,” Lee says. “But with our technique, we can actually track each individual’s complete path.” In addition to tracking the entirety of an individual monarch’s journey, this will be the first time it’s possible for conservationists to see how day-to-day environmental conditions impact their behavior. “We’ll be able to see what types of habitats they actually spend their time in,” Green says. “That will help inform where we should focus efforts for conservation activity.” Monarchs are particularly important for conservation, for they act as a sentinel species. Since monarchs travel to many different locations, they show us how the collective impact of human activities affect the wellbeing of an entire population. One of the biggest challenges has been figuring out how to pinpoint a monarch’s location, for a GPS is too large and heavy to include in the device. “We can infer the data indirectly from other primitive ultra-low power sensors using a new data-driven framework,” Kim says. The team uses deep learning algorithms and neural networks to evaluate the environmental data and infer the location based on matching conditions. The location model is created from data collected by nearly 300 volunteers who act as pseudo-butterflies. The volunteers, or citizen scientists, use sensors to collect environmental data along known monarch migration routes. “Bicyclists travel around the same speed and the same distance as monarchs do in a particular day,” Blaauw says, “so we have volunteer cyclists take larger sensors with them on multi-day trips, and we use that data to check the algorithms. It’s a bit of a role reversal, for normally we use animals to model as humans in science, but this time we’re using humans to model for animals.” “Working together with the volunteers is the most exciting part of this project,” Kim says. “It is a very rare opportunity to design an advanced machine learning algorithm using the data collected by K – 12 students and their families.” The team is hoping to do a few preliminary deployments this fall in specific local areas (COVID-19 permitting), and another deployment next fall in more distant locations. They plan to scale up gradually to full deployment that covers the entire migration range over the next year or two. This iteration of the M3 could be applied to tracking other species as well, furthering additional wildlife conservation efforts. For more information on the project or how you can volunteer, visit https://monarch.engin.umich.edu/
Hayley Hanway, ECE Communications Coordinator, University of Michigan