论文标题
测试时间培训与蒙版自动编码器
Test-Time Training with Masked Autoencoders
论文作者
论文摘要
测试时间训练通过使用自学意义的每个测试输入优化一个模型，可以随时适应新的测试分布。在本文中，我们将蒙版的自动编码器用于这个单样本学习问题。从经验上讲，我们的简单方法改善了许多视觉基准的概括，以进行分配变化。从理论上讲，我们根据偏见变化权衡取得的进步来表征这种改进。
Test-time training adapts to a new test distribution on the fly by optimizing a model for each test input using self-supervision. In this paper, we use masked autoencoders for this one-sample learning problem. Empirically, our simple method improves generalization on many visual benchmarks for distribution shifts. Theoretically, we characterize this improvement in terms of the bias-variance trade-off.
