arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1009.3589cs.LGcs.CVcs.NE
收藏

深度自学习手写体字符识别

Deep Self-Taught Learning for Handwritten Character Recognition

U. Montreal(蒙特利尔大学)
Frédéric Bastien, Yoshua Bengio, Arnaud Bergeron, Nicolas Boulanger-Lewandowski, Thomas Breuel, Youssouf Chherawala, Moustapha Cisse, Myriam Côté, Dumitru Erhan… 展开作者

Frédéric Bastien, Yoshua Bengio, Arnaud Bergeron, Nicolas Boulanger-Lewandowski, Thomas Breuel, Youssouf Chherawala, Moustapha Cisse, Myriam Côté, Dumitru Erhan, Jeremy Eustache, Xavier Glorot, Xavier Muller, Sylvain Pannetier Lebeuf, Razvan Pascanu, Salah Rifai, Francois Savard, Guillaume Sicard

摘要:

最近统计机器学习方面的理论和经验工作已经证明了学习算法对于深度架构的重要性,即,通过组合多个非线性变换获得的函数类。自学学习(利用未标记的示例或来自其他分布的示例)已经应用于深度学习器,但主要是为了展示未标记示例的优势。在这里,我们将探讨{ \em发行版外示例}带来的优势。为此,我们开发了一个强大的随机变化和噪声过程的字符图像,不仅包括仿射变换,但也倾斜,局部弹性变形,厚度的变化,背景图像,灰度变化,对比度,遮挡,和各种类型的噪声发生器。从这些高度失真的图像或通过包括与目标测试集中的对象类不同的对象类的示例来获得分布外的示例。我们证明了{ \em深度学习者比相应的浅层学习者从分布外的例子中受益更多},至少在手写字符识别领域。事实上,我们表明,他们击败了以前发表的结果,并达到人类水平的性能,手写数字分类和62类手写字符识别。

英文摘要:

Recent theoretical and empirical work in statistical machine learning has demonstrated the importance of learning algorithms for deep architectures, i.e., function classes obtained by composing multiple non-linear transformations. Self-taught learning (exploiting unlabeled examples or examples from other distributions) has already been applied to deep learners, but mostly to show the advantage of unlabeled examples. Here we explore the advantage brought by {\em out-of-distribution examples}. For this purpose we developed a powerful generator of stochastic variations and noise processes for character images, including not only affine transformations but also slant, local elastic deformations, changes in thickness, background images, grey level changes, contrast, occlusion, and various types of noise. The out-of-distribution examples are obtained from these highly distorted images or by including examples of object classes different from those in the target test set. We show that {\em deep learners benefit more from out-of-distribution examples than a corresponding shallow learner}, at least in the area of handwritten character recognition. In fact, we show that they beat previously published results and reach human-level performance on both handwritten digit classification and 62-class handwritten character recognition.

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。

↑