Inferring Neural Activity Changes under Recording Uncertainty with Transformers and Variational Inference

A difference in recorded firing rate across days may reflect a change in neuronal activity, missed spikes or recording a different neuron. We present a statistical approach that combines neuron tracking with Bayesian estimates of stationary mean recorded firing rates. Bounds under assumed differences in spike detection connect recorded rates to biological rate changes. We assess recording loss and uncertainty in rate changes in simulations, and count-model fit in recordings. We also compare two transformer estimators with Gaussian and flow variational inference under a shared temporal negative-binomial model. Flow variational inference agrees most closely with reference posteriors, while predictive performance varies across mice. Rate-change intervals contain true simulated changes less often than intended. In UnitMatch recordings from five mice, some inferred changes remain supported when spike detection is allowed to differ between days; others depend on equal detection. These interpretations require a correct neuron match. The approach identifies which inferred rate changes remain supported under specified recording assumptions and where uncertainty limits interpretation.

Authors

Publication Details

Journal
arXiv (Cornell University)
Published
2026-11-18
DOI
https://doi.org/10.48550/arxiv.2611.24767
Primary Topic
Neural dynamics and brain function
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Inferring Neural Activity Changes under Recording Uncertainty with Transformers and Variational Inference

Nilson Chapagain, Pooja Chapagain
arXiv (Cornell University)
Neural dynamics and brain function
preprint

Inferring Neural Activity Changes under Recording Uncertainty with Transformers and Variational Inference

Nilson Chapagain, Pooja Chapagain
preprint en

Abstract

A difference in recorded firing rate across days may reflect a change in neuronal activity, missed spikes or recording a different neuron. We present a statistical approach that combines neuron tracking with Bayesian estimates of stationary mean recorded firing rates. Bounds under assumed differences in spike detection connect recorded rates to biological rate changes. We assess recording loss and uncertainty in rate changes in simulations, and count-model fit in recordings. We also compare two transformer estimators with Gaussian and flow variational inference under a shared temporal negative-binomial model. Flow variational inference agrees most closely with reference posteriors, while predictive performance varies across mice. Rate-change intervals contain true simulated changes less often than intended. In UnitMatch recordings from five mice, some inferred changes remain supported when spike detection is allowed to differ between days; others depend on equal detection. These interpretations require a correct neuron match. The approach identifies which inferred rate changes remain supported under specified recording assumptions and where uncertainty limits interpretation.

arXiv (Cornell University)
Neural dynamics and brain function
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon