2025-01-14-yamaguchi25a.md

title	booktitle	year	volume	series	month	publisher	pdf	url	openreview	abstract	layout	issn	id	tex_title	firstpage	lastpage	page	order	cycles	bibtex_editor	editor	bibtex_author	author	date	address	container-title	genre	issued	extras
Analyzing Diffusion Models on Synthesizing Training Datasets	Proceedings of the 16th Asian Conference on Machine Learning	2025	260	Proceedings of Machine Learning Research	0	PMLR	https://raw.githubusercontent.com/mlresearch/v260/main/assets/yamaguchi25a/yamaguchi25a.pdf	https://proceedings.mlr.press/v260/yamaguchi25a.html	x2PY6GP9GJ	Synthetic samples from diffusion models are promising for training discriminative models as replications or augmentations of real training datasets. However, we found that the synthetic datasets degrade classification performance over real datasets when using the same dataset size. This means that the synthetic samples from modern diffusion models are less informative for training discriminative tasks. This paper investigates the gap between synthetic and real samples by analyzing the synthetic samples reconstructed from real samples through the noising (diffusion) and denoising (reverse) process of diffusion models. By varying the time steps starting the reverse process in the reconstruction, we can control the trade-off between the information in the original real data and the information produced by diffusion models. Through assessing the reconstructed samples and the trained models, we found that the synthetic samples are concentrated in modes of the training data distribution as the reverse step increases, and thus, they have difficulty covering the outer edges of the distribution. On the contrary, we found that these synthetic samples yield significant improvements in the data augmentation setting where both real and synthetic samples are used, indicating that the samples around modes are useful as interpolation for learning classification boundaries. These findings suggest that modern diffusion models are currently insufficient to replicate the real training dataset in the same dataset size but are suitable for interpolating the real training samples as the augment datasets.	inproceedings	2640-3498	yamaguchi25a	Analyzing Diffusion Models on Synthesizing Training Datasets	335	350	335-350	335	false	Nguyen, Vu and Lin, Hsuan-Tien	given family Vu Nguyen given family Hsuan-Tien Lin	Yamaguchi, Shin'ya	given family Shin’ya Yamaguchi	2025-01-14		Proceedings of the 16th Asian Conference on Machine Learning	inproceedings	date-parts 2025 1 14