Dataset Distillation with Infinitely Wide Convolutional Networks

TL;DR

This paper introduces a novel dataset distillation method using infinitely wide convolutional networks, achieving high accuracy with significantly fewer data points across multiple image classification benchmarks.

Contribution

It presents a new distributed kernel-based meta-learning framework for dataset distillation, achieving state-of-the-art results with minimal data on various datasets.

Findings

Over 65% test accuracy on CIFAR-10 with only 10 data points

Significant improvement over previous methods on multiple datasets

Preliminary analysis of distilled datasets reveals differences from natural data

Abstract

The effectiveness of machine learning algorithms arises from being able to extract useful features from large amounts of data. As model and dataset sizes increase, dataset distillation methods that compress large datasets into significantly smaller yet highly performant ones will become valuable in terms of training efficiency and useful feature extraction. To that end, we apply a novel distributed kernel based meta-learning framework to achieve state-of-the-art results for dataset distillation using infinitely wide convolutional neural networks. For instance, using only 10 datapoints (0.02% of original dataset), we obtain over 65% test accuracy on CIFAR-10 image classification task, a dramatic improvement over the previous best test accuracy of 40%. Our state-of-the-art results extend across many other settings for MNIST, Fashion-MNIST, CIFAR-10, CIFAR-100, and SVHN. Furthermore, we perform some preliminary analyses of our distilled datasets to shed light on how they differ from naturally occurring data.