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

2024-09-25

arXiv.org

AI Summary

To address the limited interpretability of CNN-based receivers in communication systems, this paper proposes an SNR-oriented neuron-level information attribution method. Grounded in information bottleneck theory, the approach quantifies neuronal sensitivity and integrates channel-parameter-guided gradient weighting with a multi-scale aggregation mechanism, establishing the first neuron attribution framework explicitly designed for communication parameters. The method enables robust estimation in high-dimensional scenarios and supports both model-level global interpretation and layer-/neuron-level local interpretation. Link-level simulations demonstrate its capability to precisely identify neurons highly sensitive to SNR variations. Ablation studies show that masking these critical neurons induces over three orders-of-magnitude increase in bit error rate, strongly validating the method’s explanatory fidelity and practical utility.

Technology Category

• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: Interpretability, Explainability, and Transparency

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Security and Privacy: Large-scale security measurements

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.

Problem

• Neural Network Models
• Signal Recognition
• Variable Sound Intensity

Innovation

• Neural Network Sensitivity Analysis
• High-Precision Estimation
• Big Data Application