A Law of Data Separation in Deep Learning
This provides a foundational insight for designing architectures, improving robustness, and interpreting predictions in AI, though it is incremental as it builds on existing understanding of neural networks.
The authors tackled the black-box nature of deep learning by studying how deep neural networks separate data by class in intermediate layers, discovering a simple quantitative law that shows each layer improves separation at a constant geometric rate across various architectures and datasets.
While deep learning has enabled significant advances in many areas of science, its black-box nature hinders architecture design for future artificial intelligence applications and interpretation for high-stakes decision makings. We addressed this issue by studying the fundamental question of how deep neural networks process data in the intermediate layers. Our finding is a simple and quantitative law that governs how deep neural networks separate data according to class membership throughout all layers for classification. This law shows that each layer improves data separation at a constant geometric rate, and its emergence is observed in a collection of network architectures and datasets during training. This law offers practical guidelines for designing architectures, improving model robustness and out-of-sample performance, as well as interpreting the predictions.