arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 0812.4044cs.LGcs.AI
收藏

部分标记学习的偏移树

The Offset Tree for Learning with Partial Labels

Yahoo Research(雅虎研究院);Microsoft Research(微软研究院);IBM Research(IBM研究院)

Alina Beygelzimer, John Langford

摘要:

我们提出了一种算法,称为偏移树,学习决策的情况下,只有一个选择的回报,而不是所有的选择。该算法将此设置减少到二进制分类,允许在此部分信息设置中重用任何现有的完全监督的二进制分类算法。我们表明,偏移树是一个最佳的减少二进制分类。特别是,它的遗憾最多是它所使用的二元分类器的遗憾的$(k-1)$倍(其中$k$是选择的数量),并且没有减少到二元分类可以做得更好。在训练和测试时,这种减少也是计算上最优的,只需要$O(\log_2 k)$工作来对示例进行训练或进行预测。实验表明,偏移树的性能一般优于其他几种方法。

英文摘要:

We present an algorithm, called the Offset Tree, for learning to make decisions in situations where the payoff of only one choice is observed, rather than all choices. The algorithm reduces this setting to binary classification, allowing one to reuse of any existing, fully supervised binary classification algorithm in this partial information setting. We show that the Offset Tree is an optimal reduction to binary classification. In particular, it has regret at most $(k-1)$ times the regret of the binary classifier it uses (where $k$ is the number of choices), and no reduction to binary classification can do better. This reduction is also computationally optimal, both at training and test time, requiring just $O(\log_2 k)$ work to train on an example or make a prediction. Experiments with the Offset Tree show that it generally performs better than several alternative 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。

↑