Uncovering the latent structure of interwoven population and temporal codes

A probabilistic latent-variable framework separates burst-related and single-spike factors in neural population data, enabling direct inference of burst coding structure from timing beyond firing rates.

• Why it matters: Existing population analyses often assume rate coding and can miss information carried by precise spike timing, especially within bursts.
• What they did: The method uses a latent-variable model to disentangle burst and spike components and is validated on both simulated and experimental datasets, including tests for burst coding presence.
• The result: Rate-only analysis obscures latent burst structure, while the framework infers the correct latent structure and supports linking burst changes to internal variables tied to attention, perception, and learning.

AI summary of the abstract · may contain errors · not medical advice · read the paper.