EigDiff: Eigensample Diffusion - or how to significantly reduce training dataset size with marginal performance variation?
Quick Answer
EigDiff introduces a novel approach to train diffusion models using eigensamples, reducing dataset size by up to 14x while maintaining performance.
Quick Take
By leveraging PCA in the latent space of a pre-trained CLIP model, the method accelerates subset selection by up to 1607x, significantly optimizing training efficiency.
Key Points
- EigDiff can reduce training data size by up to 14x.
- Subset selection accelerated by up to 1607x using PCA in CLIP space.
- Maintains model performance with marginal differences.
- Focuses on reducing data redundancy rather than diffusion steps.
- Applicable across multiple datasets and diffusion models.
Paper Resources
📖 Reader Mode
~2 min readAbstract:Diffusion models are notoriously expensive to train, requiring large-scale datasets to learn how to gradually reverse the noise from images over multiple small denoising steps. Current approaches focus on distilling the number of diffusion steps, neglecting potential redundancies of the training data. In this paper, we explore the potential of harnessing eigensamples to train diffusion models on significantly smaller datasets, while preserving model performance. To mitigate the sensitivity to affine transformations of eigenimages, we harness the latent space of a pre-trained CLIP model. We next compute latent eigenvectors by performing principal component analysis (PCA) in CLIP space. Further, we employ eigenvectors as reference points to guide the selection of a representative training set comprising comprising both exemplars (images whose embeddings are closest to eigenvectors) and diverse images. Our comprehensive experiments with different diffusion models applied on multiple datasets indicate that our approach (EigDiff) can reduce training data by up to 14x and accelerate subset selection by up to 1607x, with marginal performance differences.
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.20999 [cs.CV] |
| (or arXiv:2610.20999v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2610.20999 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Radu Tudor Ionescu [view email]
[v1]
Wed, 14 Oct 2026 22:36:03 UTC (22,925 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CV
See more →ProMoE-FL: Prototype-conditioned Mixture of Experts for Multimodal Federated Learning with Missing Modalities
ProMoE-FL introduces a Prototype-conditioned Mixture-of-Experts framework for multimodal federated learning, effectively addressing missing modalities. It outperforms existing methods on four chest X-ray datasets, demonstrating superior feature synthesis capabilities in both homogeneous and heterogeneous settings.