Events

Quantifying Variability and Individuality in Neural Population Recordings

Biomedical Engineering
Lecture / Panel
 
Open to the Public
""
Zoom

Speaker:

Alex Williams, Ph.D.
Assistant Professor, NYU Center for Neural Science
Project Leader, Flatiron Institute Center for Computational Neuroscience
Affiliate Faculty, NYU Center for Data Science, NYU Arts & Science

 

Abstract:

Quantifying differences across species and individuals is fundamental to many fields of biology. However, it remains challenging to rigorously compare large-scale recordings of neural populations across different animals or brain regions. In this talk, Dr. Williams will introduce approaches that formalize such comparisons using metric spaces—spaces in which one can define distances that are symmetric and satisfy the triangle inequality. These formal constructions enable powerful downstream analyses, including clustering and nearest-neighbor prediction. He will discuss applications spanning multiple behavioral tasks, such as navigation, passive visual processing, and decision making, and across species, including mice and non-human primates, highlighting the potential of these methods to connect population-level neural geometry to behavior.

Dr. Williams leads the Laboratory for Neural Statistics, which develops statistical models and open-source computational tools to extract insight from large-scale neural data. His research focuses on characterizing flexibility and variability in neural circuits and on how population dynamics change during learning, attention, development, and aging, all of which remain difficult to summarize even at a descriptive level. His prior work has introduced tensor decomposition models for trial-by-trial gain modulation, time-warping approaches for temporal variability, and methods for neural sequence discovery using convolutional and Bayesian nonparametric techniques.

Dr. Williams earned a BA in Neuroscience (summa cum laude) with a minor in Computer Science from Bowdoin College. He received his PhD in Neurosciences from Stanford University in 2019, where he worked with Surya Ganguli, and subsequently completed postdoctoral training in the Stanford Department of Statistics with Scott Linderman. In 2022, he joined the faculty of the Center for Neural Science at NYU. His research has been recognized with a McKnight Scholar Award (2023–2026) and is supported by multiple NIH BRAIN Initiative grants, including projects focused on scalable neurostatistical software and computational methods to characterize variability in large-scale neural circuits.

Shape Metric Analysis Diagram
Shape Metric Analysis of Simulated 1D Tuning Curve: Top: four example datasets (out of K = 40) of N = 100 neurons sorted by their preferred stimulus θ, with varying tuning width and concentration of tuning around θ = 0. Bottom: the first 3 principal components of the same datasets show how tuning curve parameters shape neural manifolds.