arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2103.16591cs.LGcs.AI
收藏

持续重量平衡

Continuous Weight Balancing

Stanford University(斯坦福大学)

Daniel J. Wu, Avoy Datta

摘要:

我们提出了一个简单的方法,通过它来选择样本权重的问题,高度不平衡或倾斜的性状。我们不是天真地离散回归标签来找到分组权重,而是采取一种更有原则的方法-我们从估计的源分布和指定的目标分布之间的传递函数导出样本权重。我们的方法在回归和分类任务上都优于未加权和离散加权模型。我们还向科学界开源了我们对这种方法的实现(https://github.com/Daniel-Wu/Continuous-Weight-Balancing)。

英文摘要:

We propose a simple method by which to choose sample weights for problems with highly imbalanced or skewed traits. Rather than naively discretizing regression labels to find binned weights, we take a more principled approach -- we derive sample weights from the transfer function between an estimated source and specified target distributions. Our method outperforms both unweighted and discretely-weighted models on both regression and classification tasks. We also open-source our implementation of this method (https://github.com/Daniel-Wu/Continuous-Weight-Balancing) to the scientific community.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑