Coordinate-VAE: Unsupervised clustering and de-noising of peripheral nervous system data

Thomas J Hardcastle, Susannah Lee, Lorenz Wernisch, Pascal Fortier-Poisson, Sudha Shunmugam, Kalon Hewage, Tris Edwards, Oliver Armitage, Emil Hewage

TL;DR: Unsupervised analysis of data recorded from the peripheral nervous system denoises and categorises signals.

Keywords: Machine Learning, Peripheral Nervous System, Convolutional Neural Networks, Auto-encoder, Signal Processing

Abstract: The peripheral nervous system represents the input/output system for the brain. Cuff electrodes implanted on the peripheral nervous system allow observation and control over this system, however, the data produced by these electrodes have a low signal-to-noise ratio and a complex signal content. In this paper, we consider the analysis of neural data recorded from the vagus nerve in animal models, and develop an unsupervised learner based on convolutional neural networks that is able to simultaneously de-noise and cluster regions of the data by signal content.