Google Research

KIP Distilled Datasets


These are distilled datasets derived from MNIST, Fashion-MNIST, CIFAR-10, CIFAR-100, and SVHN using infinitely wide convolutional networks. Sample result: Over 64% test accuracy on CIFAR-10 achieved using only 10 images. Datasets distilled are using the KIP (Kernel Inducing Points) and LS (Label Solve) algorithms studied in Dataset Metalearning from Kernel Ridge-Regression and Dataset Distillation with Infinitely Wide Convolutional Networks.