Precision Learning: Towards Use of Known Operators in Neural Networks

TL;DR

This paper explores integrating known physical transforms into neural networks to improve accuracy, demonstrated through X-ray material decomposition, reducing errors and enhancing prediction quality by leveraging prior knowledge.

Contribution

It introduces a method to incorporate known transforms into neural networks, reducing error bounds and improving prediction accuracy in physics-based applications.

Findings

Inclusion of known transforms reduces maximal error bounds.

Application to X-ray decomposition improves SSIM from 0.54 to 0.88.

Method applicable to various physics and signal processing tasks.

Abstract

In this paper, we consider the use of prior knowledge within neural networks. In particular, we investigate the effect of a known transform within the mapping from input data space to the output domain. We demonstrate that use of known transforms is able to change maximal error bounds. In order to explore the effect further, we consider the problem of X-ray material decomposition as an example to incorporate additional prior knowledge. We demonstrate that inclusion of a non-linear function known from the physical properties of the system is able to reduce prediction errors therewith improving prediction quality from SSIM values of 0.54 to 0.88. This approach is applicable to a wide set of applications in physics and signal processing that provide prior knowledge on such transforms. Also maximal error estimation and network understanding could be facilitated within the context of precision learning.