Mesoscopic modeling of hidden spiking neurons
This work addresses the challenge of underconstrained modeling in neuroscience for researchers studying large-scale neural recordings, though it is incremental as it builds on existing latent variable models with a more transparent biological mapping.
The authors tackled the problem of modeling unobserved neurons in spiking neural networks (SNNs) for generative modeling of multi-neuronal recordings, by developing a neuronally-grounded latent variable model (neuLVM) that uses coarse-graining and mean-field approximations to reduce hidden neuron activity to a low-dimensional mesoscopic description, enabling efficient model inversion to recover connectivity parameters and infer latent population activity from a few observed neurons.
Can we use spiking neural networks (SNN) as generative models of multi-neuronal recordings, while taking into account that most neurons are unobserved? Modeling the unobserved neurons with large pools of hidden spiking neurons leads to severely underconstrained problems that are hard to tackle with maximum likelihood estimation. In this work, we use coarse-graining and mean-field approximations to derive a bottom-up, neuronally-grounded latent variable model (neuLVM), where the activity of the unobserved neurons is reduced to a low-dimensional mesoscopic description. In contrast to previous latent variable models, neuLVM can be explicitly mapped to a recurrent, multi-population SNN, giving it a transparent biological interpretation. We show, on synthetic spike trains, that a few observed neurons are sufficient for neuLVM to perform efficient model inversion of large SNNs, in the sense that it can recover connectivity parameters, infer single-trial latent population activity, reproduce ongoing metastable dynamics, and generalize when subjected to perturbations mimicking photo-stimulation.