Taming Diffusion for Dataset Distillation with High Representativeness
This work addresses the need for compact, cost-efficient datasets in deep learning, offering an incremental improvement over existing diffusion-based methods for dataset distillation.
The paper tackles the problem of dataset distillation by addressing issues in current diffusion-based methods, such as inaccurate distribution matching and distribution deviation, and proposes D^3HR, a framework that achieves higher accuracy across different model architectures compared to state-of-the-art baselines.
Recent deep learning models demand larger datasets, driving the need for dataset distillation to create compact, cost-efficient datasets while maintaining performance. Due to the powerful image generation capability of diffusion, it has been introduced to this field for generating distilled images. In this paper, we systematically investigate issues present in current diffusion-based dataset distillation methods, including inaccurate distribution matching, distribution deviation with random noise, and separate sampling. Building on this, we propose D^3HR, a novel diffusion-based framework to generate distilled datasets with high representativeness. Specifically, we adopt DDIM inversion to map the latents of the full dataset from a low-normality latent domain to a high-normality Gaussian domain, preserving information and ensuring structural consistency to generate representative latents for the distilled dataset. Furthermore, we propose an efficient sampling scheme to better align the representative latents with the high-normality Gaussian distribution. Our comprehensive experiments demonstrate that D^3HR can achieve higher accuracy across different model architectures compared with state-of-the-art baselines in dataset distillation. Source code: https://github.com/lin-zhao-resoLve/D3HR.