A team of 麻豆原创 researchers placed eighth in the 2025 EEG Challenge, a global machine learning competition that asks participants to predict behavioral responses from brain data. The Knights, who call themselves Team Marque, bested 8,400 submissions, including those from research labs and tech companies like Meta and Emotiv.
The winning team includes Mubarak Shah, the director of the 麻豆原创 Institute of Artificial Intelligence (IAI); Helen Huang and Qiushi Fu, associate professors of biomedical engineering; Yue Wen, an assistant professor of biomedical engineering; Abhilash Durgam, a doctoral student who works in the Center for Research in Computer Vision; and Jerry Fu, a postdoctoral scholar mentored by Huang and Wen.
As top 10 winners, Team Marque鈥檚 code will be added to the competition鈥檚 open-source repository, contributing to the future advancement of EEG research. They also receive a certificate in recognition of their achievement. Shah says that placing in the top 10 at the world鈥檚 premier venue for AI and machine learning is a tremendous accomplishment for 麻豆原创 and its newly established IAI.
鈥淚t speaks to the strength of 麻豆原创鈥檚 interdisciplinary culture,鈥� Shah says.
鈥淥ur students and faculty, with their combined expertise in machine learning, neuroscience, signal processing and computer vision can compete with some of the world鈥檚 best teams.鈥� 鈥� Mubarak Shah, Trustee Chair Professor
The competitors had to prevail in two individual challenges that utilized data from the Healthy Brain Network, which includes EEGs of more than 3,000 children who were multitasking. Challenge 1 asked the teams to improve the predicted reaction time of a subject seeing change in contrast of an image while Challenge 2 called for an improved prediction of mental health traits in a subject.
Durgam says the secret to Team Marque鈥檚 success was to look for the patterns that hold true for all people.
鈥淩ather than treat this as a regression problem to predict a number, we used a classification approach where we taught our model to recognize the unique ‘profile’ of the person,鈥� Durgam says. 鈥淭his encouraged the model to understand the individual’s distinct characteristics rather than just treating the task as a simple math problem.鈥�
The team鈥檚 efforts are more than just an accomplishment for themselves and for the university 鈥� their code can now be used by scientists to advance EEG research.
鈥淥ur open-source repository supports open-science efforts, which I believe is necessary to make substantial breakthroughs in EEG research at a faster rate than any one group could accomplish alone,鈥� Huang says. 鈥淏eing able to predict mental health traits in developing children is a challenging problem that has great societal impact and could be solved faster collectively as a field by working in parallel and sharing data and code so groups don鈥檛 have to repeat something that has already been tried.鈥�
Team Marque came together after Durgam reached out to Huang to learn more about EEG. Each of them had already formed teams for the competition, but decided to combine efforts for better results. For Huang, the competition also had a personal connection as one of the organizers, Seyed Yahya Shirazi 鈥�21PhD, is her former student.
鈥淚 don鈥檛 think we have been in the top 10 if we didn鈥檛 combine efforts,鈥� Huang says. 鈥淭ogether, we could work in parallel to explore fundamentally different approaches first to identify the most promising one and then focus on optimizing specific parameters.鈥�
