Diversity-Driven Synthesis: Enhancing Dataset Distillation through Directed Weight Adjustment

Overview

Dataset distillation aims to compress a large dataset into a much smaller synthetic dataset while preserving its training utility. Our method, Directed Weight Adjustment (DWA), enhances dataset distillation by promoting diversity in the synthetic data through carefully directed adjustments to the optimization process.

This paper was accepted as a Spotlight presentation at NeurIPS 2024.

Links

• Paper: arXiv:2409.17612
• Code: GitHub Repository

Citation

@inproceedings{dwa2024neurips,
    title={Diversity-Driven Synthesis: Enhancing Dataset Distillation through Directed Weight Adjustment},
    author={Du, Jiawei and Zhang, Xin and Hu, Juncheng and Huang, Wenxin and Zhou, Joey Tianyi},
    booktitle={Adv. Neural Inf. Process. Syst. (NeurIPS)},
    year={2024}
}