DREAM: Efficient Dataset Distillation by Representative Matching

TL;DR

DREAM introduces a novel representative matching strategy for dataset distillation, selecting key original images to improve efficiency and performance, reducing training iterations significantly without sacrificing quality.

Contribution

The paper proposes DREAM, a new matching strategy that selects representative images for dataset distillation, enhancing efficiency and achieving state-of-the-art results.

Findings

Reduces distilling iterations by over 8 times.

Achieves state-of-the-art performance with sufficient training.

Improves stability and diversity in dataset synthesis.

Abstract

Dataset distillation aims to synthesize small datasets with little information loss from original large-scale ones for reducing storage and training costs. Recent state-of-the-art methods mainly constrain the sample synthesis process by matching synthetic images and the original ones regarding gradients, embedding distributions, or training trajectories. Although there are various matching objectives, currently the strategy for selecting original images is limited to naive random sampling. We argue that random sampling overlooks the evenness of the selected sample distribution, which may result in noisy or biased matching targets. Besides, the sample diversity is also not constrained by random sampling. These factors together lead to optimization instability in the distilling process and degrade the training efficiency. Accordingly, we propose a novel matching strategy named as \textbf{D}ataset distillation by \textbf{RE}present\textbf{A}tive \textbf{M}atching (DREAM), where only representative original images are selected for matching. DREAM is able to be easily plugged into popular dataset distillation frameworks and reduce the distilling iterations by more than 8 times without performance drop. Given sufficient training time, DREAM further provides significant improvements and achieves state-of-the-art performances.