从故事的一面了解真相
Learning the Truth From Only One Side of the Story
摘要：
在单侧反馈下学习（即，我们只观察我们积极预测的例子的标签）是机器学习中的一个基本问题--应用包括贷款和推荐系统。尽管如此，令人惊讶的是，在减轻抽样偏差影响方面取得的进展甚微。我们专注于广义线性模型，并表明，如果不调整这种采样偏差，该模型可能会收敛次优，甚至无法收敛到最优解。我们提出了一种自适应的方法，来与理论保证，并表明它优于现有的方法经验。我们的方法利用方差估计技术来有效地在不确定性下学习，与现有方法相比，提供了一种更有原则的替代方案。
英文摘要：
Learning under one-sided feedback (i.e., where we only observe the labels for examples we predicted positively on) is a fundamental problem in machine learning -- applications include lending and recommendation systems. Despite this, there has been surprisingly little progress made in ways to mitigate the effects of the sampling bias that arises. We focus on generalized linear models and show that without adjusting for this sampling bias, the model may converge suboptimally or even fail to converge to the optimal solution. We propose an adaptive approach that comes with theoretical guarantees and show that it outperforms several existing methods empirically. Our method leverages variance estimation techniques to efficiently learn under uncertainty, offering a more principled alternative compared to existing approaches.
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。
