Curvature-Aware Zeroth-Order Optimization for Memory-Efficient Test-Time Adaptation
面向内存高效测试时适应的曲率感知零阶优化
机构 * Shanghai Jiao Tong University(上海交通大学)
AI总结 本文针对零阶测试时适应方法梯度估计方差高的问题,提出CAZO方法,利用损失海森结构优化扰动采样,在降内存开销的同时实现了最优跨域适应性能。
Comments Published in CVPR 2026
Journal ref Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2026, pp. 836-846