样本价值相对学习者而定:冻结子集干预揭示学习器依赖
Get the story
2026-10-02,arXiv Machine Learning Theory 发表一手研究,提出“样本价值相对学习者而定”的观点,并通过冻结子集干预实验加以验证:在低分辨率 ImageNet-100 上,将 ResNet-18 宽度加倍后,easy-first 与几何覆盖的交叉边界从 57 移至 85(每类样本);扩大输入网格而保持图像信息不变时边界左移,且宽度增大会削弱该左移。研究据此指出同一样本对不同学习者的相对价值会随模型结构变化。
Generated from reports · updated 4 hr ago
Timeline
Follow the coverage from different angles.
- arXiv · Machine Learning TheoryUseful to Whom? Sample Value Is Defined Only Relative to the Learner
summary_zh 冻结子集干预实验表明,同一样本对不同学习者的相对价值会随模型结构变化:在低分辨率 ImageNet-100 上,将 ResNet-18 宽度加倍使 easy-first 与几何覆盖的交叉边界从 57 移至 85 每类样本;扩大输入网格而保持图像信息不变时边界左移,且宽度增大会削弱该左移。
Heat trend
Current heat 9·Comparable peak 10(Oct 2)·Comparable change over 24 hours –
The trend compares only the same participants observed continuously; its range may be smaller than the current heat count. Move or click on the chart to inspect hourly heat; use the left and right arrow keys to switch.