English

Data-driven Regularization via Racecar Training for Generalizing Neural Networks

Computer Vision and Pattern Recognition 2020-07-02 v1

Abstract

We propose a novel training approach for improving the generalization in neural networks. We show that in contrast to regular constraints for orthogonality, our approach represents a {\em data-dependent} orthogonality constraint, and is closely related to singular value decompositions of the weight matrices. We also show how our formulation is easy to realize in practical network architectures via a reverse pass, which aims for reconstructing the full sequence of internal states of the network. Despite being a surprisingly simple change, we demonstrate that this forward-backward training approach, which we refer to as {\em racecar} training, leads to significantly more generic features being extracted from a given data set. Networks trained with our approach show more balanced mutual information between input and output throughout all layers, yield improved explainability and, exhibit improved performance for a variety of tasks and task transfers.

Keywords

Cite

@article{arxiv.2007.00024,
  title  = {Data-driven Regularization via Racecar Training for Generalizing Neural Networks},
  author = {You Xie and Nils Thuerey},
  journal= {arXiv preprint arXiv:2007.00024},
  year   = {2020}
}

Comments

https://github.com/tum-pbs/racecar

R2 v1 2026-06-23T16:44:51.140Z