Inversion dynamics of class manifolds in deep learning reveals tradeoffs underlying generalisation

arXiv:2303.05161v2 Announce Type: replace
Abstract: To achieve near-zero training error in a classification problem, the layers of a feed-forward network have to disentangle the manifolds of data points with different labels, to facilitate the discrimination. However, excessive class separation can bring to overfitting since good generalisation requires learning invariant features, which involve some level of entanglement. We report on numerical experiments showing how the optimisation dynamics finds representations that balance these opposing tendencies with a non-monotonic trend. After a fast …

abstract arxiv class classification cs.lg data deep learning discrimination dynamics error good labels near network overfitting training type