Learning to cluster neuronal function
Nina S. Nellen, Polina Turishcheva, Michaela Vystrčilová, Shashwat Sridhar, Tim Gollisch, Andreas S. Tolias, Alexander S. Ecker · 2025
Read Learning to cluster neuronal function by Nina S. Nellen, Polina Turishcheva, Michaela Vystrčilová, Shashwat Sridhar, Tim Gollisch, Andreas S. Tolias, Alexander S. Ecker (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-03293.txt.
Description
Nina Nellen and co-authors teach neural models of the visual cortex to sort neurons into tidy functional families, in a 2025 paper. Networks trained to predict brain activity from images form digital twins of mouse visual cortex, and each neuron gets an embedding, a fingerprint vector. Those fingerprints should cluster by cell type, but in practice they smear.
Their method, DECEMber, adds a clustering pressure to training: a loss borrowed from t-distributions pulls embeddings toward cluster centers, which an expectation-maximization routine updates alongside the network. The result keeps prediction accuracy high while making the clusters far more consistent than post-hoc clustering achieves. Whether mouse cortex truly lacks tidy types remains open, but blurred tools are no longer the excuse.
Catalog record
| CREATOR | Nina S. Nellen, Polina Turishcheva, Michaela Vystrčilová, Shashwat Sridhar, Tim Gollisch, Andreas S. Tolias, Alexander S. Ecker |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 36 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-03293