Dataset Meta-Learning from Kernel Ridge-Regression

TL;DR

This paper introduces Kernel Inducing Points (KIP), a meta-learning algorithm that creates significantly smaller or corrupted datasets maintaining model performance, advancing dataset distillation and privacy-preserving applications.

Contribution

The paper presents KIP, a novel meta-learning method inspired by kernel ridge regression, achieving superior dataset compression and transferability for neural network training.

Findings

KIP compresses datasets by one or two orders of magnitude.

KIP achieves state-of-the-art results on MNIST and CIFAR-10.

KIP-learned datasets transfer effectively to finite-width neural networks.

Abstract

One of the most fundamental aspects of any machine learning algorithm is the training data used by the algorithm. We introduce the novel concept of $\epsilon$-approximation of datasets, obtaining datasets which are much smaller than or are significant corruptions of the original training data while maintaining similar model performance. We introduce a meta-learning algorithm called Kernel Inducing Points (KIP) for obtaining such remarkable datasets, inspired by the recent developments in the correspondence between infinitely-wide neural networks and kernel ridge-regression (KRR). For KRR tasks, we demonstrate that KIP can compress datasets by one or two orders of magnitude, significantly improving previous dataset distillation and subset selection methods while obtaining state of the art results for MNIST and CIFAR-10 classification. Furthermore, our KIP-learned datasets are transferable to the training of finite-width neural networks even beyond the lazy-training regime, which leads to state of the art results for neural network dataset distillation with potential applications to privacy-preservation.