Abstract

In this journal club, we will explore how infinite hidden semi-markov models, which combine Bayesian nonparametrics with hidden Markov model-like dynamics, can be used for automatic segmentation of behaviours during learning. We will begin by providing an overview of infinite hidden Semi-Markov models [2], and contrasting them with the classical infinite hidden Markov model [1]. Then, extending this framework to a behavioural modelling context, we will present a recent nature neuroscience paper by Bruins et al. [3], which uses an infinite hidden semi-Markov model to capture the learning dynamics of individual mice in a large dataset as they performed a perceptual decision making task. By allowing for modest variations in the existing states, as well as the introduction of new states throughout the task, they demonstrate three main phases of learning across the population of mice, alongside large inter-individual differences. Overall, we aim to provide a view into Bayesian nonparametrics as a valuable tool for automatic state segmentation and for the characterisation of individual learning trajectories in a neuroscientific context.

[1] Beal, M. J., Ghahramani, Z., & Rasmussen, C. E. (2001). The infinite hidden Markov model. Advances in Neural Information Processing Systems, 14. [2] Johnson, M. J., & Willsky, A. S. (2013). Bayesian nonparametric hidden semi-Markov models. Journal of Machine Learning Research, 14, 673–701. [3] Bruijns, S. A., Bougrova, K., Laranjeira, I. C., et al. (2026). Infinite hidden Markov models can dissect the complexities of learning. Nature Neuroscience, 29(1), 186–194.