Neural population coding
How populations represent sensory, behavioral, and task variables; how structure is distributed across neurons; and how sensitivity, variability, and covariance shape what can be inferred from population activity.
PhD Student · Stephenson School of Biomedical Engineering · University of Oklahoma
I work on neural population coding and the geometry of neural representations.
I am especially interested in mathematical and statistical tools for studying population activity, trial-to-trial variability, and geometric structure in neural recordings and neural network models.
amirhossein.yavari (at) ou.edu
I am a PhD student in the Stephenson School of Biomedical Engineering at the University of Oklahoma, advised by Prof. Farnaz Zamani Esfahlani. Prior to this, I completed my undergraduate studies in Mathematics and its Applications at Sharif University of Technology.
My research interests are centered on neural population coding and interpretable structure in population activity. I am interested in how groups of neurons represent sensory, behavioral, and task variables; how structure in those responses is organized across a population; and how sensitivity, variability, and response geometry constrain what can be inferred from neural activity.
I am also interested in how coding properties vary across brain areas, subjects, stimulus families, model architectures, and time. This direction connects with broader work in our group on brain dynamics, functional organization, and network-level structure, especially in settings where the stability of population-level measurements matters.
How populations represent sensory, behavioral, and task variables; how structure is distributed across neurons; and how sensitivity, variability, and covariance shape what can be inferred from population activity.
How neural codes change across trials, sessions, learning, context, and internal state, and when population-level measurements remain stable enough to support comparison across time.
How the geometry of population responses relates to sensitivity, invariance, ambiguity, and decodability, and how geometric descriptions can clarify what is being compared across systems.
Loading publications…
Preprint posted: Beyond Activation Alignment: The Geometry of Neural Sensitivity .
Presented at SfN 2025: Dynamics of Cortico-Subcortical Interactions in Functional Brain Networks .
Third place, American Airlines Operation Research and Advanced Analytics Hackathon.
Started PhD at the University of Oklahoma.