Stop memorizing: A data-dependent regularization framework for intrinsic pattern learning

TL;DR

This paper introduces GRSVNet, a new data-dependent regularization framework that aligns feature geometry with loss functions, enabling neural networks to learn intrinsic patterns and resist memorizing random data.

Contribution

The paper proposes GRSVNet, a novel regularization framework that enforces geometry during training and validation, improving discriminative features and preventing memorization of noise.

Findings

OLE-GRSVNet outperforms conventional regularized DNNs on real data.

OLE-GRSVNet refuses to memorize random data or labels.

The framework enhances learning of intrinsic patterns by reducing memorization capacity.

Abstract

Deep neural networks (DNNs) typically have enough capacity to fit random data by brute force even when conventional data-dependent regularizations focusing on the geometry of the features are imposed. We find out that the reason for this is the inconsistency between the enforced geometry and the standard softmax cross entropy loss. To resolve this, we propose a new framework for data-dependent DNN regularization, the Geometrically-Regularized-Self-Validating neural Networks (GRSVNet). During training, the geometry enforced on one batch of features is simultaneously validated on a separate batch using a validation loss consistent with the geometry. We study a particular case of GRSVNet, the Orthogonal-Low-rank Embedding (OLE)-GRSVNet, which is capable of producing highly discriminative features residing in orthogonal low-rank subspaces. Numerical experiments show that OLE-GRSVNet outperforms DNNs with conventional regularization when trained on real data. More importantly, unlike conventional DNNs, OLE-GRSVNet refuses to memorize random data or random labels, suggesting it only learns intrinsic patterns by reducing the memorizing capacity of the baseline DNN.