A3D3 Seminar: Lu Mi

Monday 14 Apr 2025, 12:00 → 13:15 US/Pacific

Title: Accelerating Scientific Discovery in the Brain in the Age of AI

 

Abstract: 

Neuroscience is entering an era where artificial intelligence drives new discoveries by unraveling complex brain data at unprecedented scales. This talk explores how we could use advanced AI models to analyze multi-modal brain datasets encompassing neural activity, connectivity, behavior, and transcriptomics. Such models open opportunities to integrate diverse data streams and reveal patterns underlying brain function, but they also pose challenges: how do we ensure these AI models remain scalable to massive datasets, interpretable to neuroscientists, and generalizable across experiments and subjects? We will discuss emerging strategies to address these challenges, moving beyond black-box predictions towards models that yield mechanistic insights into neural systems.

Key examples include NetFormer, an interpretable transformer architecture that captures dynamic neural connectivity and plasticity, and a biophysics-constrained variational autoencoder (VAE) that integrates physical constraints to uncover latent structure in neural data. We also highlight active acquisition frameworks for closed-loop experimental design, where AI systems propose informative new experiments or data collection strategies to accelerate learning. Additionally, the talk will introduce advanced tools for automated brain data collection at nanoscale resolution—intelligent microscopy and high-throughput pipelines that pair with AI models to map neural circuits with unprecedented detail. By integrating these cutting-edge approaches, the talk will illustrate a path toward scalable, generalizable models that not only predict complex brain phenomena but also guide experimental science and uncover the underlying principles of brain computation.

 

 

Lu Mi comes to Georgia Tech from the Allen Institute for Brain Science and the University of Washington where she was a Shanahan Foundation Fellow. She completed her Ph.D. at MIT in 2022 with Professor Nir Shavit. Mi’s research lies at the intersection of natural intelligence and artificial intelligence. She mainly focuses on building fast and scalable automatic brain imaging pipelines. By leveraging artificial neural networks, Mi aims to understand the mechanisms of coding, computation, and learning within biological neural networks, using multi-modal neural data. She is also interested in developing brain-inspired machine intelligence frameworks.

 

The A3D3 Seminar is a monthly lecture series that hosts scholars working across applied areas of artificial intelligence, such as hardware algorithm co-development, high energy physics, multi-messenger astrophysics,  and neuroscience. Our presenters come from all four domain fields and include occasional external speakers beyond the A3D3 science areas, governmental agencies and industry. The seminar will be recorded and published in YouTube. To receive future event updates, subscribe here.