Unsupervised Neural Universal Denoiser for Finite-Input General-Output Noisy Channel

TL;DR

This paper introduces Gen-CUDE, an unsupervised neural network denoiser for finite-input, general-output channels that outperforms existing methods in denoising accuracy and computational efficiency.

Contribution

It presents a novel neural network-based universal denoiser that is unsupervised, computationally efficient, and theoretically supported, applicable to a broad class of noisy channels.

Findings

Gen-CUDE achieves better denoising results than strong baselines.

It has a tighter theoretical upper bound on performance.

The method works well on both synthetic and real data.

Abstract

We devise a novel neural network-based universal denoiser for the finite-input, general-output (FIGO) channel. Based on the assumption of known noisy channel densities, which is realistic in many practical scenarios, we train the network such that it can denoise as well as the best sliding window denoiser for any given underlying clean source data. Our algorithm, dubbed as Generalized CUDE (Gen-CUDE), enjoys several desirable properties; it can be trained in an unsupervised manner (solely based on the noisy observation data), has much smaller computational complexity compared to the previously developed universal denoiser for the same setting, and has much tighter upper bound on the denoising performance, which is obtained by a theoretical analysis. In our experiments, we show such tighter upper bound is also realized in practice by showing that Gen-CUDE achieves much better denoising results compared to other strong baselines for both synthetic and real underlying clean sequences.