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 1409.5834cs.LGcs.DSstat.ML
收藏

结构化预测的紧误差界

Tight Error Bounds for Structured Prediction

Amir Globerson, Tim Roughgarden, David Sontag, Cafer Yildirim

摘要:

机器学习中的结构化预测任务涉及多个标签的同时预测。这通常通过最大化标签空间上的得分函数来完成,该得分函数分解为成对元素的总和,每个元素取决于两个特定的标签。直观地说,使用的成对项越多,预期的准确性就越好。然而,目前还没有关于这种直觉的理论解释。本文件朝着这个方向迈出了重要的一步。我们制定的问题分类的顶点的一个已知的图$G=(V,E)$,其中的顶点和边的图形标记和相关的半随机与地面真相。我们表明,实现低预期的汉明误差的前景取决于有趣的方式图$G$的结构。例如,如果$G$是非常差的扩展器,如路径,则大的预期汉明误差是不可避免的。我们的主要积极的结果表明,对于广泛的一类图形,包括在机器视觉应用中常见的2D网格图,有一个多项式时间的算法与小和信息理论上接近最优的预期误差。我们的研究结果提供了第一步的理论依据的经验成功的有效近似推理算法,用于结构化预测模型,精确的推理是棘手的。

英文摘要:

Structured prediction tasks in machine learning involve the simultaneous prediction of multiple labels. This is typically done by maximizing a score function on the space of labels, which decomposes as a sum of pairwise elements, each depending on two specific labels. Intuitively, the more pairwise terms are used, the better the expected accuracy. However, there is currently no theoretical account of this intuition. This paper takes a significant step in this direction. We formulate the problem as classifying the vertices of a known graph $G=(V,E)$, where the vertices and edges of the graph are labelled and correlate semi-randomly with the ground truth. We show that the prospects for achieving low expected Hamming error depend on the structure of the graph $G$ in interesting ways. For example, if $G$ is a very poor expander, like a path, then large expected Hamming error is inevitable. Our main positive result shows that, for a wide class of graphs including 2D grid graphs common in machine vision applications, there is a polynomial-time algorithm with small and information-theoretically near-optimal expected error. Our results provide a first step toward a theoretical justification for the empirical success of the efficient approximate inference algorithms that are used for structured prediction in models where exact inference is intractable.

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。

↑