Very interesting. A related paper from a couple of years ago proposed a similar idea to understand generalization in deep learning: https://arxiv.org/abs/2203.

Very interesting. A related paper from a couple of years ago proposed a similar idea to understand generalization in deep learning: https://arxiv.org/abs/2203.10036 https://arxiv.org/abs/2203.10036