#BayesianModels

Fabrizio MusacchioFabMusacchio
2026-01-06

🧠 New paper by Bruijns et al.: Infinite hidden models can dissect the complexities of learning

They introduce a dynamic infinite that treats as a sequence of discrete, reusable behavioral states rather than a smooth learning curve. Applied to >100 mice, it reveals robust learning stages despite massive individual variability and shows that early biases do not predict later skill.

🌍 doi.org/10.1038/s41593-025-021

Fig. 1 | Behavior and modeling overview.

Client Info

Server: https://mastodon.social
Version: 2025.07
Repository: https://github.com/cyevgeniy/lmst