1. Brain Computer Interface

For the first exchangemeet, we met with Siwei Mai from Integrated Systems & NeuroImaging Laboratory led by Prof. Laleh Najafizadeh.

A little background on how it happened: Siwei and I were both TA for the Digital System Design. I got to know that he reserach was in brain computer interface (BCI) and showed me one of his resaerch poster. During my undergrad (in 2014), I was planning to build some applications based on BCI for undergrad thesis but couldn’t do it as I could not acquire the required hardware. When Siwei mentioned his work in BCI, I was thrilled and immediately thought his lab would be an ideal place to initiate the ExchangeMeet.

Through our discussions, we delved into various aspects of BCI, including its fundamentals, challenges, potential applications, and overlaps with our research interests. A key area of convergence we identified was graph representation, which plays a significant role in both fields. In BCI, graphs are used to model brain neuron connections, known as the connectome, providing insights into the structure and function of neural networks. Advancing graph representation techniques is critical for understanding the connectome, while my lab focuses on efficiently extracting information from large-scale graphs.

Another shared area of interest is matrix factorization, where our lab has explored graph-based methods to accelerate computation, a technique that aligns with BCI’s data analysis and signal processing needs.

We also discussed emerging trends in BCI, particularly the integration of large language models (LLMs). With the growing excitement around LLMs, it’s intriguing to explore what a foundational model could mean for different domains. In the context of BCI, one promising application is the development of a foundational model capable of generalizing the analysis of EEG signals across various channels and configurations. Such a model could greatly enhance the scalability and adaptability of BCI systems, paving the way for more robust and versatile applications.