Paper Notes: LRP Optimization for Background Bias

Posted on Edited on In Views:
Paper: Improving deep neural network generalization and robustness to background bias via layer-wise relevance propagation optimization. [PDF]

Overview

Background and Motivation

Key Contributions

Methodology

Layer-wise Relevance Propagation (LRP)

Optimization Strategy

Experiments

Key Takeaways

My Understanding

Conclusion

This is a reading note from my study of interpretable machine learning methods.