简体繁体 English

使用条件随机场的多标签分类

[英]MultiLabel Classification using Conditional Random Field

原文 2016-05-30 18:11:31 1 2 python/ machine-learning/ classification/ multilabel-classification/ crf

Is it possible to use Conditional Random Field for MultiLabel Classification?是否可以将条件随机字段用于多标签分类？ I saw a python CRF implementation at https://pystruct.github.io/user_guide.html , but couldn't figure a way to do multilabel classification.我在https://pystruct.github.io/user_guide.html看到了一个 python CRF 实现，但无法找到进行多标签分类的方法。

2 个解决方案

The basic CRF doesn't support multilabel classification.基本 CRF 不支持多标签分类。 However, some extensions have been explored, such as the Collective Multi-label (CML) and the Collective Multi-label with Features (CMLF).但是，已经探索了一些扩展，例如集体多标签（CML）和具有特征的集体多标签（CMLF）。 From (1):从（1）：

A conditional random field (CRF) based model is presented in [21] where two multi-label graphical models has been proposed, both parameterizes label co-occurances.在 [21] 中提出了一个基于条件随机场 (CRF) 的模型，其中提出了两个多标签图形模型，它们都将标签共现参数化。 The Collective Multi-label (CML) classifier maintains feature accounting for label co-occurances and the Collective Multi-label with Features (CMLF) maintains parameters that correspond to features for each co-occuring label pair.集体多标签 (CML) 分类器维护标签共现的特征，而带特征的集体多标签 (CMLF) 维护与每个共现标签对的特征相对应的参数。 Petterson et.彼得森等。 al.阿尔。 recently presented another interesting generative modeling approach in a reverse manner, predicting a set of instances given the labels [39].最近以相反的方式提出了另一种有趣的生成建模方法，预测给定标签的一组实例 [39]。

References:参考资料：

(1) Sorower, Mohammad S. "A literature survey on algorithms for multi-label learning." (1) Sorower, Mohammad S.“关于多标签学习算法的文献调查。” Oregon State University, Corvallis (2010).俄勒冈州立大学，科瓦利斯 (2010)。 http://people.oregonstate.edu/~sorowerm/pdf/Qual-Multilabel-Shahed-CompleteVersion.pdf ; http://people.oregonstate.edu/~sorowerm/pdf/Qual-Multilabel-Shahed-CompleteVersion.pdf ； https://scholar.google.com/scholar?cluster=11211211207326445005&hl=en&as_sdt=0,22 https://scholar.google.com/scholar?cluster=11211211207326445005&hl=en&as_sdt=0,22
(21) N. Ghamrawi and A. Mccallum. (21) N. Ghamrawi 和 A. Mccallum。 Collective Multi-Label Classification.集体多标签分类。 In Proceedings of the 3005 ACM Conference on Information and Knowledge Management (CIKM '05), pages 195–200, Bremen, Germany, 2005. http://www.dtic.mil/dtic/tr/fulltext/u2/a440081.pdf在 3005 ACM 信息和知识管理会议 (CIKM '05) 的会议记录中，第 195-200 页，德国不来梅，2005 年。http: //www.dtic.mil/dtic/tr/fulltext/u2/a440081.pdf
(39) James Petterson and Tiberio Caetano. (39) James Petterson 和 Tiberio Caetano。 Reverse multi-label learning.逆向多标签学习。 In J. Lafferty, CKI Williams, R. Zemel, J. Shawe-Taylor, and A. Culotta, editors, Advances in Neural Information Processing Systems 23, pages 1903–1911.在 J. Lafferty、CKI Williams、R. Zemel、J. Shawe-Taylor 和 A. Culotta，编辑，神经信息处理系统进展 23，第 1903-1911 页。 2010. 2010 年。

I encountered a modified CRF named fuzzy CRF as shown below.我遇到了一个名为模糊 CRF的修改后的CRF ，如下所示。

Its mathematics is quite simple as we can see from equation 2 in the paper:从论文中的方程 2 可以看出，它的数学非常简单：

We just sum all the energies of the paths in the numerator, and the denominator remains the same.我们只是将分子中路径的所有能量相加，分母保持不变。 For inference, we can apply Viterbi or beam search.对于推理，我们可以应用维特比或波束搜索。