arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1205.6432cs.LG
收藏

多类学习方法:理论比较与启示

Multiclass Learning Approaches: A Theoretical Comparison with Implications

The Hebrew University(希伯来大学)

Amit Daniely, Sivan Sabato, Shai Shalev Shwartz

摘要:

我们从理论上分析和比较了以下五种流行的多类分类方法:One vs. All,All Pairs,Tree-based classifiers,带有随机生成代码矩阵的纠错输出代码(ECOC)和MulticlassSVM。在前四种方法中,分类是基于二进制分类的简化。我们考虑的情况下,二元分类器来自一类VC维$d$,特别是从类的半空间在$\reals^d$。我们分析了这些方法的估计误差和近似误差。我们的分析揭示了有趣的结论的实际意义,在各种条件下的不同方法的成功。我们的证明技术采用VC理论的工具来分析假设类的\emph{approximation error}。这与大多数(如果不是全部的话)以前使用VC理论形成鲜明对比,这些理论只处理估计误差。

英文摘要:

We theoretically analyze and compare the following five popular multiclass classification methods: One vs. All, All Pairs, Tree-based classifiers, Error Correcting Output Codes (ECOC) with randomly generated code matrices, and Multiclass SVM. In the first four methods, the classification is based on a reduction to binary classification. We consider the case where the binary classifier comes from a class of VC dimension $d$, and in particular from the class of halfspaces over $\reals^d$. We analyze both the estimation error and the approximation error of these methods. Our analysis reveals interesting conclusions of practical relevance, regarding the success of the different approaches under various conditions. Our proof technique employs tools from VC theory to analyze the \emph{approximation error} of hypothesis classes. This is in sharp contrast to most, if not all, previous uses of VC theory, which only deal with estimation error.

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。

↑