Title - In vitro living neurons for computation through dynamic control and reservoir computing
Abstract - Engineering a living system for computing gained popular attention, motivated by their ability to efficiently construct complex circuits for cognitive functions through networks of neurons. Moreover, the energy efficiency inherent in spike-based neuromorphic computing holds promising potential: processing continuous streams of multi-modal sensory inputs and coordinating actuation seems prevalent in biological systems while remaining an aspiration for engineers and roboticists. One unavoidable challenge in working with biological systems is their continuous, ever-changing dynamics, which can limit scalability and complicate the development of a robust computational framework. In this talk, we examine the behavior of living neuronal cultures to characterize their dynamics and develop strategies to overcome these challenges. Insights from the reservoir computing (RC) literature help in understanding the mechanism of processing information with rich dynamical systems. Importantly, the performance of an optimal dynamical system must have an intrinsic balance of non-linearity in temporal dynamics. We introduce a dynamic control method that fine-tunes this balance, achieving the necessary consistency and separation properties as benchmarked by RC. The demonstration is carried out on MiV-OH experimental platform containing in-vitro motoneuron cultures derived from optogenetic stem cells, interfaced with micro-electrode arrays (MEA) to capture their electrical activity. With several years of accumulated data, we highlight the significance of establishing optimal tuning to maximize not only computational ability but also the durability of the trained model over the system's living span. This study advances our technical understanding of neuromorphic computing using biological substrates and offers crucial insights into overcoming practical limitations.
Bio - Seung Hyun Kim is a 5th-year Ph.D. student in Theoretical and applied mechanics, advised by Prof. Mattia Gazzola. Seung has been working on the Mind in Vitro project to study the potential of utilizing biological neural substrates for neuromorphic computing.
Part of the Illinois Computer Science Speakers Series.
Food will be provided after the seminar.
If accommodation is required, please email <communications@cs.illinois.edu>. Someone from our staff will contact you to discuss your specific needs.