CVJun 30, 2020

Data-driven Regularization via Racecar Training for Generalizing Neural Networks

arXiv:2007.00024v1
Originality Incremental advance
AI Analysis

This work addresses generalization issues in neural networks for machine learning practitioners, though it appears incremental as it builds on existing orthogonality constraints with a data-driven twist.

The authors tackled the problem of improving neural network generalization by introducing a data-dependent orthogonality constraint, which led to more generic features, improved explainability, and enhanced performance across various tasks and transfers.

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.

Code Implementations1 repo
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes