Dataset distillation improves by matching sample difficulty precisely
PSM: Dataset Distillation Based on Precise Statistical Matching by Difficulty
Computer Vision and Pattern Recognition
Summary
Large datasets are often compressed into smaller ones to train AI faster, but existing methods miss how some samples are harder than others. The authors propose a new approach that ranks samples by difficulty and matches these groups when creating small datasets. This tailored method leads to better performance on different tasks. Their experiments show it works across many types of data and AI models.
What this means in practice
- •For machine learning engineers: Create smaller training datasets that preserve complexity variations for quicker model training with less data.
- •For computer vision developers: Improve model accuracy by distilling image datasets using difficulty-aware techniques tailored to specific image groups.
Authors
Hongxu Ma, Guang Li, Shijie Wang, Dongzhan Zhou, Suorong Yang, Baoli Sun, Takahiro Ogawa, Miki Haseyama, Zhihui Wang
Abstract
Dataset distillation (DD) condenses a large original dataset into a small distilled dataset with high training utility. Decoupled statistical matching methods substantially reduce distillation time and memory overhead while achieving strong performance. However, they typically supervise all distilled samples using running statistics estimated from the entire original dataset. These statistics mainly capture the average feature distribution while overlooking differences in sample difficulty, limiting their ability to characterize the difficulty structure of the original data. To address this issue, we propose Precise Statistical Matching (PSM) by difficulty. After pretraining, PSM uses the Global Precision Score (GPS) to estimate image difficulty, ranks the samples within each class, and partitions each class into IPC (images per class) difficulty groups. During distillation, Statistics Updated Again (SUA) updates the teacher's batch normalization (BN) running statistics through forward passes on original samples from each group, providing difficulty-specific supervision for the corresponding distilled batch. Meanwhile, Initial Sample Screening (ISS) initializes distilled samples using original images from the corresponding difficulty group, providing an effective starting point for precise matching. Experiments across multiple datasets and model architectures demonstrate that PSM broadens the difficulty range of distilled samples and improves downstream performance in most evaluated settings. Code will be released.