Penn State's Institute of Computational and Data Sciences awarded 14 projects that will use data science and computational power to investigate some of science's biggest challenges and opportunities.
Seed grants to fund projects that tackle huge scientific, societal challenges
Posted on July 22, 2022UNIVERSITY PARK, Pa. — Penn State’s Institute for Computational and Data Sciences awarded this year’s seed grant program, offering more funds to help Penn State scientists kickstart research projects.
The seed grants will fund 14 different proposals from researchers who represent five different colleges and 11 different departments, including several interdisciplinary collaborations that stretch across colleges.
Some of the seed grants will fund carbon capture technology, ultrasonics, galaxy surveys, artificial intelligence, equitable disease prediction and quantum machine learning. Here are this year’s funded projects:
Carbon Capture
Carbon capture is a technology that is capturing the interest of scientists, who believe it might be a tool to tackle climate change. Yun Kyung Shin, associate research professor of mechanical engineering, will lead efforts to use seed grant funds to provide a computer-guided design of better carbon capture systems using multi-scale approaches, from ReaxFF atomistic-scale simulations to machine-learning models. Carbon capture is seen as a key technology in the fight against climate change.
Additive Manufacturing
The study of materials in the additive manufacturing of metals, ceramics, composites and cement-like materials is increasingly important to a range of industries, from medical to aerospace. Parisa Shokouhi, associate professor of engineering science and mechanics, will serve as primary investigator in a seed grant-backed project to build a digital ultrasonics laboratory using a transfer learning approach in characterizing the microstructural properties of materials, such as grain-sized and pore-sized distributions.
Galaxy Survey Moonshot
Computers are helping astronomers map the vastness of space, prompting new discoveries, as well as inspiring new questions. Joel Leja and Victoria Ashley Villar, both assistant professors of astronomy and astrophysics and both ICDS co-hires, have been awarded seed grant funding for to use a machine learning technique that promises to be about 10 million times faster than traditional analyses to better interpret galaxy imaging.
Zentropy Theory
Scientists suggest that using a theory they call zentropy theory, which predicts how temperature changes can impact volume at a range of scales, could radically change the way materials are designed, especially ones exposed to extremely high temperatures. Yi Wang, research professor in the Department of Materials Science and Engineering, will lead a team on an ICDS seed grant-funded project that will use data science to investigate zentropy theory for predictive modeling in materials research.
Communication-Efficient Distributed Machine Learning
Interest in artificial intelligence methods generally referred to as distributed learning and federated learning has grown in recent years, but communication costs in the networks processing these methods continue to be a bottleneck. Jinghui Chen, assistant professor of information sciences and technology, was awarded funding for the project, “On Communication-Efficient Adaptive Gradient Method for Large-Scale Distributed Learning,” to help reduce the communication cost for training large scale machine learning models.
Quantum Machine Learning
Quantum computing, which uses the principles of quantum mechanics to make calculations, is an emerging technology that promises to unleash massive computational power on certain tasks, including machine learning. Xiantao Li, professor of mathematics in Penn State’s Eberly College of Science, will serve as the primary investigator to explore quantum computer-based machine-learning, specifically in the use of quantum computing algorithms to accelerate the training of machine-learning models.
AI for Better Medical Engineering
Polymers are important in biomedical applications, such as tissue engineering, and medical devices, such as stents and catheters. The researchers plan to use machine learning techniques as part of a method to better predict design parameters of polymeric systems used in bio applications. Dundar Yilmaz, associate research professor in the Department of Mechanical Engineering, will serve as primary investigator for an ICDS seed grant-funded project that will explore the use AI-driven multiscale simulations to study polymer biodegradation, which could lead to improvements in these medical applications and devices.
Thermodynamic Deep Learning
Thermodynamic free energy can predict phenomena ranging from protein folding to the structure of the universe. Performing calculations of free energy in soft materials currently is complex and can take a long time. Wesley Reinhart, assistant professor of materials science and engineering and ICDS co-hire, received a seed grant to create a machine learning model that can be used to compute free energy in soft materials and then will be benchmarked against conventional methods. If successful, the AI-based modeling strategy could accelerate scientific innovation in a wide range of application areas including computational biology, chemistry and materials science.
AI Cybersecurity Agents
Control-flow hijacking — where attackers overwrite program control data to affect the victim’s control flow and eventually hijack the program — was once the main method for hackers to compromise computer systems. Fortunately, protections have been deployed against those attacks. Hackers are now exploring data-oriented exploits that change non-control data to revive their malicious capabilities. To defend against these attacks, cybersecurity professionals need to identify non-control, security-critical data, which often relies on human efforts. Hong Hu, assistant professor of information sciences and technology and ICDS affiliate, will serve as primary investigator for a cybersecurity project that seeks to identify non-control, security-critical data through deep learning techniques in defense against these attacks.
Safe, Abundant Renewable Energy
Chalcogenide perovskites could serve as stable, non-toxic and abundant materials for use in photovoltaics, a promising renewable energy technology used in solar cells. Nelson Dzade, assistant professor of energy and mineral engineering and ICDS associate, was awarded and will serve as the primary investigator on a project to better understand the design of these materials, which could then be used in clean energy products.
Equitable Disease Prediction
Developing novel statistical methods to integrate various genomic data sets may pave the way for more equitable — and more effective — disease prediction in diverse human populations. Xiang Zhu, assistant professor of statistics and ICDS associate, will serve as the primary investigator for the project, “Powerful and equitable disease prediction from large-scale DNA biobanks and single-cell multiomics.”
Climate Change and the Ocean
The ocean isn’t just an inert blob of water. The overturning circulation in the ocean actively moves heat poleward, and thus scientists liken it to a “global ocean conveyor belt.” The system has important implications for climate prediction and climate change science. A research team, headed by Laifang Li, assistant professor of meteorology and atmospheric science and ICDS co-hire, was awarded a seed grant for a project titled “Learning to Reconstruct the Ocean’s Great Conveyor Belt.”
Powering Electric Vehicles
More electric vehicles are hitting the road and the skies. Making these vehicles more efficient could help these sustainable modes of transportation compete with carbon-emitting vehicles. Herschel Pangborn, assistant professor of mechanical engineering, and Daning Huang, assistant professor of aerospace engineering, received a seed grant to use deep learning to promote transformative improvements in the efficiency, performance and safety of electro-thermal systems used for electrified transportation.
Examining Risks of Adult Stimulant Use
Off-label use of prescription stimulants are mainly used to treat Attention-Deficit Hyperactivity Disorder (ADHD) in children and adolescent populations, has been rapidly increasing in the adult population. However, there is little information on the risk of adverse health events associated with stimulant use and the individual-level risk factors for adults. An ICDS seed grant will fund a project, led by Qiushi Chen, assistant professor of industrial and manufacturing engineering, that will leverage large real-world electronic health record data to systematically examine individual risk and population impact of prescription stimulant use in this adults.
You can find out more about the ICDS seed grant program here.