标题： 测试时的相关性对齐 显示英文标题

标题： Test-time Correlation Alignment

摘要： 深度神经网络在分布偏移下经常退化。尽管领域适应提供了一种解决方案，但隐私约束通常会阻止访问源数据，使得仅使用未标记测试数据进行测试时适应（TTA）变得越来越有吸引力。然而，当前的 TTA 方法仍然面临实际挑战：(1) 主要关注实例级对齐，由于缺少源相关性而忽略了 CORrelation ALignment (CORAL)；(2) 复杂的反向传播操作用于模型更新，导致计算开销和(3) 领域遗忘。为了解决这些挑战，我们提供了理论分析以研究测试时相关性对齐 (TCA) 的可行性，证明了高置信度实例与测试实例之间的相关性对齐可以在理论上保证提升测试性能。基于此，我们提出了两种简单但有效的算法：LinearTCA 和 LinearTCA+。LinearTCA 通过简单的线性变换实现实例和相关性对齐，而无需额外的模型更新，而 LinearTCA+ 可作为即插即用模块轻松提升现有的 TTA 方法。广泛的实验验证了我们的理论洞察，并表明 TCA 方法在各种任务、基准和主干上显著优于基线。值得注意的是，与最佳 TTA 基线相比，LinearTCA 在仅使用 4% 的 GPU 内存和 0.6% 的计算时间的情况下实现了更高的准确性。它还在 CLIP 上比现有方法高出 1.86%。 显示英文摘要

摘要： Deep neural networks often degrade under distribution shifts. Although domain adaptation offers a solution, privacy constraints often prevent access to source data, making Test-Time Adaptation (TTA, which adapts using only unlabeled test data) increasingly attractive. However, current TTA methods still face practical challenges: (1) a primary focus on instance-wise alignment, overlooking CORrelation ALignment (CORAL) due to missing source correlations; (2) complex backpropagation operations for model updating, resulting in overhead computation and (3) domain forgetting. To address these challenges, we provide a theoretical analysis to investigate the feasibility of Test-time Correlation Alignment (TCA), demonstrating that correlation alignment between high-certainty instances and test instances can enhance test performances with a theoretical guarantee. Based on this, we propose two simple yet effective algorithms: LinearTCA and LinearTCA+. LinearTCA applies a simple linear transformation to achieve both instance and correlation alignment without additional model updates, while LinearTCA+ serves as a plug-and-play module that can easily boost existing TTA methods. Extensive experiments validate our theoretical insights and show that TCA methods significantly outperforms baselines across various tasks, benchmarks and backbones. Notably, LinearTCA achieves higher accuracy with only 4% GPU memory and 0.6% computation time compared to the best TTA baseline. It also outperforms existing methods on CLIP over 1.86%.