Mara Ellison Voss
Postdoctoral Researcher · Zürich, Switzerland
I build models of how populations of neurons encode uncertainty — and I write, in public, about the messy parts of doing that work.
I study the geometry of neural population codes: how distributed activity across thousands of neurons represents belief, uncertainty, and time. My work sits between dynamical-systems theory and modern representation learning, and I care about models that are both predictive and interpretable.
Before Zürich I completed my PhD at the Gatsby Computational Neuroscience Unit, where I worked on latent variable models for large-scale electrophysiology. I still write most of my analysis code in the open.
News
- "The geometry of uncertainty" is out in Nature Neuroscience.
- Awarded an SNSF Spark grant for uncertainty benchmarking.
- Talk at the Bernstein Conference on population geometry.
Selected publicationsall publications →
Recent writingAll posts →
Why I think uncertainty lives on curved manifolds, and a small experiment you can run in an afternoon to see it for yourself.
Publishing this blog from Notion with elogThe exact pipeline behind this site — a Notion database, elog, and a front-matter contract I fully control.
Three papers that changed how I read data this springA short annotated reading list — dimensionality, double descent, and the quiet return of the humble GLM.