SODA: Spectral Orthogonal Decomposition Adaptation for Diffusion Models.

• Tác giả: Ligong Han, Junzhou Huang, Felix Juefei-Xu, Dimitris Metaxas, Vladimir Pavlovic, Akash Srivastava, Molei Tao, Hao Wang, Song Wen, Xinxi Zhang
• Ngôn ngữ: eng
• Ký hiệu phân loại: 616.845 Convulsions
• Thông tin xuất bản: United States : IEEE Winter Conference on Applications of Computer Vision. IEEE Winter Conference on Applications of Computer Vision , 2025
• Bộ sưu tập: NCBI
• ID: 748339

Tóm tắt

Adapting large-scale pre-trained generative models in a parameter-efficient manner is gaining traction. Traditional methods like low rank adaptation achieve parameter efficiency by imposing constraints but may not be optimal for tasks requiring high representation capacity. We propose a novel spectrum-aware adaptation framework for generative models. Our method adjusts both singular values and their basis vectors of pretrained weights. Using the Kronecker product and efficient Stiefel optimizers, we achieve parameter-efficient adaptation of orthogonal matrices. Specifically, we introduce