A Bayesian nonparametric approach for uncovering rat hippocampal population codes during spatial navigation

Area of Science:

• Neuroscience
• Computational Neuroscience
• Machine Learning

Background:

• Rodent hippocampal population codes are crucial for spatial navigation.
• Existing computational methods aim to decode spatial topology from neural activity.

Purpose of the Study:

• To develop a novel Bayesian nonparametric approach for inferring rat hippocampal population codes during navigation.
• To address model selection challenges in neural decoding.

Main Methods:

• Application of a hierarchical Dirichlet process-hidden Markov model (HDP-HMM).
• Utilized Bayesian inference methods: Markov chain Monte Carlo (MCMC) and Variational Bayes (VB).
• Demonstrated effectiveness on freely behaving rat navigation data.

Main Results:

• The HDP-HMM significantly outperforms traditional finite-state HMMs on simulated and experimental data.
• MCMC-based inference with Hamiltonian Monte Carlo (HMC) sampling proved flexible and efficient.
• The proposed Bayesian methods surpassed VB and empirical Bayes approaches for hyperparameter setting.

Conclusions:

• The Bayesian nonparametric HDP-HMM enables efficient model selection and parameter identification.
• This method is suitable for modeling latent-state neuronal population dynamics.
• Advances neural decoding of spatial navigation from hippocampal ensembles.