Explainable Neural Network-based Modulation Classification via Concept\n Bottleneck Models

2021/01/04 by Lauren J. Wong, Wong, Lauren J., Sean McPherson +1 · 1 citation
Computer Science · #FOS: Electrical engineering #Signal Processing (eess.SP) #Wireless Signal Modulation Classification #electronic engineering #information engineering

paper · pdf · doi:10.48550/arxiv.2101.01239

openalex publication_date 2021/01/04 · openalex created_date 2022/07/25 · openalex updated_date 2026/07/28

Abstract

While RFML is expected to be a key enabler of future wireless standards, a\nsignificant challenge to the widespread adoption of RFML techniques is the lack\nof explainability in deep learning models. This work investigates the use of CB\nmodels as a means to provide inherent decision explanations in the context of\nDL-based AMC. Results show that the proposed approach not only meets the\nperformance of single-network DL-based AMC algorithms, but provides the desired\nmodel explainability and shows potential for classifying modulation schemes not\nseen during training (i.e. zero-shot learning).\n

Cited by

• Towards Interpretable PolSAR Image Classification: Polarimetric Scattering Mechanism Informed Concept Bottleneck and Kolmogorov-Arnold Network

Related

• DL-AMC: Deep Learning for Automatic Modulation Classification
• Using Early Exits for Fast Inference in Automatic Modulation Classification
• Deep Neural Network Architectures for Modulation Classification
• Edge-Efficient Deep Learning Models for Automatic Modulation Classification: A Performance Analysis
• Deep Modulation Embedding