Background

Modern network routing has become increasingly complex, with ML models making critical routing decisions. However, these models often act as black boxes, making it difficult to understand their decision-making process. GradCAM (Gradient-weighted Class Activation Mapping) has emerged as a powerful visualization technique in deep learning, offering insights into model decisions through heat map visualizations. Applying GradCAM to network routing can provide network operators with visual explanations of routing decisions, enhancing trust and enabling better network management.

Problem Description

This thesis focuses on adapting GradCAM for network routing applications. The project consists of the following steps:

Acquired Knowledge and Skills

Through this thesis, you will gain expertise in explainable AI techniques, deep learning architectures, and network routing principles. You will work with modern visualization tools and develop skills in implementing interpretability methods for ML models in networking contexts.