Dependable modulation classifier explainer with measurable explainability

Area of Science:

• Telecommunications Engineering
• Artificial Intelligence in Networks
• Signal Processing

Background:

• Smart cities rely on the Internet of Things (IoT) for enhanced services, with IoT devices depending on robust networks.
• Artificial Intelligence (AI) optimizes network performance, reduces costs, and enables new telecom services for IoT.
• Reliable communication in IoT necessitates robust networks and effective modulation classification for signal transmission.

Purpose of the Study:

• To propose a dependable modulation classifier explainer (DMCE) for deep learning-based adaptive modulation classification (AMC).
• To enhance the explainability of modulation classification within communication systems.
• To provide methods for visualizing and numerically measuring the interpretability of AI predictions in signal modulation.

Main Methods:

• Utilized deep learning methods for adaptive modulation classification (AMC).
• Developed a Dependable Modulation Classifier Explainer (DMCE) to provide insights into classification predictions.
• Implemented data point highlighting for visual explanation and an Explainability Measurable Metric (EMM) for quantitative interpretation.

Main Results:

• Demonstrated the ability to visualize key data points influencing modulation class predictions.
• Introduced a numeric metric (EMM) for quantifying the explainability of modulation classification.
• Presented a comparative analysis showcasing the effectiveness of the proposed explainability methods against existing techniques.

Conclusions:

• The proposed DMCE enhances the transparency and trustworthiness of AI-driven modulation classification in communication systems.
• Explainable AI methods are crucial for understanding and validating predictions in critical network applications like smart cities.
• The developed techniques offer a significant advancement in the interpretability of adaptive modulation classification.