Xinhao Zhong, Shuoyang Sun, Xulin Gu, Zhaoyang Xu, Yaowei Wang, Min Zhang, Bin Chen $^{\dagger}$

ArXiv preprint arXiv:2412.09959

We propose a novel framework orthogonal to existing diffusion-based dataset distillation methods, leveraging diffusion models for selection rather than generation.

Xinhao Zhong $^{*}$, Hao Fang $^{*}$, Bin Chen $^{\dagger}$, Xulin Gu, Meikang Qiu, Shuhan Qi, Shu-Tao Xia

In The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2025

we propose a novel parameterization method dubbed Hierarchical Parameterization Distillation (H-PD), to systematically explore hierarchical feature within provided feature space (e.g., layers within pre-trained generative adversarial networks).

Xinhao Zhong, Bin Chen $^{\dagger}$, Hao Fang, Xulin Gu, Shu-Tao Xia, En-Hui Yang

In The International Conference on Learning Representations (ICLR), 2025

we introduce conditional mutual information (CMI) to assess the class-aware complexity of a dataset and propose a novel method by minimizing CMI.