Variational Joint Filtering (VJF) is a flexible framework that online learns latent nonlinear state dynamics and filters latent states from high-dimensional spike trains.
VJF is amenable to real-time applications, enables experimentalist to monitor complex data in the ongoing experiments at an abstract level, and has the potential to automate analysis and experimental design in ways that testably track and modify behavior using stimuli designed to influence learning.
Zhao, Y. and Park, I.M. Variational online learning of neural dynamics. Frontiers in Computational Neuroscience, 2020.