Bayesian nonparametric methods for discovering latent structures of rat hippocampal ensemble spikes

Hippocampal functions are responsible for encoding spatial and temporal dimensions of episodic memory, and hippocampal reactivation of previous awake experiences in sleep is important for learning and memory consolidation. Therefore, uncovering neural representations of hippocampal ensemble spike activity during various behavioral states would provide improved understanding of neural mechanisms of hippocampal-cortical circuits. In this paper, we propose two Bayesian nonparametric methods for this purpose: the Bayesian modeling allows to impose informative priors and constraints into the model, whereas Bayesian nonparametrics allows automatic model selection. We validate these methods to three different hippocampal ensemble recordings under different task behaviors, and provide interpretation and discussion on the derived results.