Multi-View Networks for Denoising of Arbitrary Numbers of Channels

TL;DR

This paper introduces multi-view neural networks that can denoise data from any number of channels at runtime, outperforming traditional models and generalizing to unseen channel configurations.

Contribution

The paper presents novel multi-view network architectures capable of handling arbitrary channel counts and leveraging additional recordings for improved denoising performance.

Findings

Outperform traditional denoising models in multi-channel scenarios

Can operate with an arbitrary number of channels at runtime

Generalize to unseen numbers of recordings during training

Abstract

We propose a set of denoising neural networks capable of operating on an arbitrary number of channels at runtime, irrespective of how many channels they were trained on. We coin the proposed models multi-view networks since they operate using multiple views of the same data. We explore two such architectures and show how they outperform traditional denoising models in multi-channel scenarios. Additionally, we demonstrate how multi-view networks can leverage information provided by additional recordings to make better predictions, and how they are able to generalize to a number of recordings not seen in training.