Interpreting Deep Neural Network-Based Receiver Under Varying Signal-To-Noise Ratios

TL;DR

This paper introduces a new interpretability method for CNN-based radio receivers, revealing which units encode key channel information across varying SNRs, aiding understanding and robustness.

Contribution

It presents a novel interpretability approach for neural network-based receivers, applicable to high-dimensional data and generalizable beyond radio applications.

Findings

Effectively identifies units linked to SNR processing.

Provides global and local explanations of neural network behavior.

Demonstrates robustness in high-dimensional settings.

Abstract

We propose a novel method for interpreting neural networks, focusing on convolutional neural network-based receiver model. The method identifies which unit or units of the model contain most (or least) information about the channel parameter(s) of the interest, providing insights at both global and local levels -- with global explanations aggregating local ones. Experiments on link-level simulations demonstrate the method's effectiveness in identifying units that contribute most (and least) to signal-to-noise ratio processing. Although we focus on a radio receiver model, the method generalizes to other neural network architectures and applications, offering robust estimation even in high-dimensional settings.