6/23/2021 Aaron Seidlitz, Illinois CS
Among the best ways to advance medical applications of dynamic brain connectivity, according to Illinois CS professor Sanmi Koyejo, is via spatiotemporal analysis. Its complexities, though, require addressing fundamental issues, which he is prepared to understand and develop.
Written by Aaron Seidlitz, Illinois CS
Illinois CS professor Sanmi Koyejo is optimistic about the advancements that machine learning models can provide medical practitioners. But first, new tools are needed to address existing roadblocks and weaknesses.
Koyejo’s latest project funded through the National Science Foundation – a five-year NSF CAREER Award for $625,000 – seeks to build on his previous work and further establish “Probabilistic Models for Spatiotemporal Data with Applications to Dynamic Brain Connectivity.”
This project will investigate three challenges Koyejo has identified in machine learning for high-dimensional spatiotemporal data: modeling flexibility, scalability, and mitigating data biases.
“We address modeling bias using a variety of scientifically motivated model components,” Koyejo said. “We can improve data bias using new distributed and federated learning approaches that enable joint training across multiple sites without explicit data sharing. Finally, this research will improve scalability using new fast probabilistic inference techniques.”
The work will influence Koyejo’s undergraduate course offerings and provide new research opportunities for graduate students from underrepresented groups. Together, these groups will all work through the challenges, bolstered by the desire to capture this technology’s practical impact on the neurosciences.
“Within neuroscience, this research will result in new tools to explain the source of individual fluctuations in brain connectivity. Such advances in science are an important step towards developing novel biomarkers for mental health disorders,” Koyejo said.
Koyejo believes the funding from NSF will allow for his progression in two primary ways. First, he can develop collaborative partnerships on and off campus. Second, it marks further validation that his efforts advance fundamental and applied research in machine learning and neuroscience.
“This award builds on the hard work of the excellent students and postdocs in my lab, along with external collaborators – particularly at the Beckman Institute,” Koyejo said. “Beyond the science that this grant will support, this award is recognition by my peers and the NSF of this research agenda, which I find humbling. I am excited to push this research forward.”
Finally, Koyejo is proud to extend the opportunity for this research work to underrepresented graduate students – an effort he remains dedicated to in his role with Illinois CS.
“Science benefits when we ensure the inclusion of a broad diversity of voices with different backgrounds,” Koyejo said. “I have followed through on this belief with my outreach efforts, some of which this award will support. In this project, the students will learn and contribute to a wide variety of topics, particularly in neuroscience and machine learning.”