Class Interference of Deep Neural Networks
This addresses a fundamental challenge in deep learning for improving model robustness and accuracy, though it appears incremental as it builds on existing error analysis concepts.
The paper identifies class interference as a key source of generalization errors in deep neural networks, proposing methods like cross-class tests and interference models to analyze and detect it during training.
Recognizing and telling similar objects apart is even hard for human beings. In this paper, we show that there is a phenomenon of class interference with all deep neural networks. Class interference represents the learning difficulty in data, and it constitutes the largest percentage of generalization errors by deep networks. To understand class interference, we propose cross-class tests, class ego directions and interference models. We show how to use these definitions to study minima flatness and class interference of a trained model. We also show how to detect class interference during training through label dancing pattern and class dancing notes.