Practical Dataset Distillation Based on Deep Support Vectors

TL;DR

This paper proposes a practical dataset distillation method that leverages Deep Support Vectors and Deep KKT loss to improve performance when only limited data is available, reducing computational demands.

Contribution

It introduces a novel distillation approach using Deep Support Vectors and Deep KKT loss, suitable for scenarios with partial dataset access.

Findings

Improved performance over baseline on CIFAR-10

Deep Support Vectors provide unique beneficial information

Enhanced distillation results with the proposed method

Abstract

Conventional dataset distillation requires significant computational resources and assumes access to the entire dataset, an assumption impractical as it presumes all data resides on a central server. In this paper, we focus on dataset distillation in practical scenarios with access to only a fraction of the entire dataset. We introduce a novel distillation method that augments the conventional process by incorporating general model knowledge via the addition of Deep KKT (DKKT) loss. In practical settings, our approach showed improved performance compared to the baseline distribution matching distillation method on the CIFAR-10 dataset. Additionally, we present experimental evidence that Deep Support Vectors (DSVs) offer unique information to the original distillation, and their integration results in enhanced performance.