LFADS - Latent Factor Analysis via Dynamical Systems

TL;DR

LFADS is a novel variational auto-encoder based method that infers low-dimensional latent dynamics from high-dimensional neural spiking data, improving analysis of large-scale neural recordings.

Contribution

Introduces LFADS, a dynamical systems-based variational auto-encoder for extracting latent neural dynamics from single-trial neural data.

Findings

LFADS outperforms existing methods in inferring neural firing rates.

LFADS accurately recovers latent dynamics from synthetic data.

The method effectively reduces high-dimensional neural data to low-dimensional factors.

Abstract

Neuroscience is experiencing a data revolution in which many hundreds or thousands of neurons are recorded simultaneously. Currently, there is little consensus on how such data should be analyzed. Here we introduce LFADS (Latent Factor Analysis via Dynamical Systems), a method to infer latent dynamics from simultaneously recorded, single-trial, high-dimensional neural spiking data. LFADS is a sequential model based on a variational auto-encoder. By making a dynamical systems hypothesis regarding the generation of the observed data, LFADS reduces observed spiking to a set of low-dimensional temporal factors, per-trial initial conditions, and inferred inputs. We compare LFADS to existing methods on synthetic data and show that it significantly out-performs them in inferring neural firing rates and latent dynamics.