# 关于Multi-label Classification 多标签分类的问题

$\Chi$多分类：

$\chi$为训练样本集，$y=\{1,\cdots,k\}$为标签集合，给定一组形式为$$的训练样本，$x_i \in \chi$$Y_i \in 2^{|y|}$，目的是学习求出合适的误差低的函数$f(x)$$2^{|y|}$中的unique values）。

$rank_f(x,l)$$f$规则下样本$x$所对应的标签$l$的排序。其中，$rank$是一个到$\{1,\cdots,k\}$上的一对一的映射，且如果$f(x,l_1) \leq f(x,l_2)$，那么$rank_f(x,l_1) \leq rank_f(x,l_2)$

Problem Transformation Method 问题转化

At training time, with $D$:

1 Transform the multi-label training data to single-label data

2 Learn from the single-label transformed data

At testing time, for $\tilde{x}$:

1 Make single-label predictions

2 Translate these into multi-label predictions

e.g.

Binary Relevance (BR): $L$ binary problems (one vs. all)

2 Label Powerset (LP): one multi-class problem of $2^L$ class-values 二进制

3 Pairwise (PW): $\frac{L(L-1)}{2}$ binary problems (all vs. all)

Each model is trained based on examples annotated by at least one of the labels, but not both.

4 Copy-Weight (CW): one multi-class problem of L class values

Make a single multi-class problem with L possible class values.Each example duplicated $|Y^{(i)})|$times, weighted as$\frac{1}{|Y^{(i)}|}$.

Algorithm Independent 独立于算法

Label-based Transformation 基于标签的转化

Instance-based Transformation 基于实例的转化

Instance Elimination

Label Creation

Conversion

Label Elimination (Simplification)

Labels Decomposition

Multiplicative

Algorithm Independent Method总结

Algorithm Adaptation / Dependent Method 算法适应

e.g.,

k-Nearest Neighbours (KNN)

MLkNN [Zhang and Zhou, 2007]

Decision Trees (DT)

Multi-label C4.5 [Clare and King, 2001]

Support Vector Machines (SVM)

RankSVM, a Maximum Margin approach [Elisseeff and Weston, 2002]

Godbole and Sarawagi (2004) 结合PT4

Aiolli, F. & Sperduti, A. (2005) Multiclass Classification with Multi-Prototype Support Vector Machines.

Neural Networks (NN)

BPMLL [Zhang and Zhou, 2006]

An extensive empirical study by [Madjarov et al., 2012] recommends:
RT-PCT: Random Forest of Predictive Clustering Trees (Algorithm Adaptation, Decision Tree based)
HOMER: Hierarchy Of Multilabel ClassifiERs (Problem Transformation, LP-based (original presentation))
CC: Classifier Chains (Problem Transformation, BR-based)

André C. P. L. F. de Carvalho, Freitas A A. A Tutorial on Multi-label Classification Techniques[M]// Foundations of Computational Intelligence Volume 5. Springer Berlin Heidelberg, 2009:177-195.

Li T, Zhang C, Zhu S. Empirical Studies on Multi-label Classification.[C]// IEEE International Conference on TOOLS with Artificial Intelligence. IEEE Computer Society, 2006:86-92.

Multi-Label Classification: An Overview

©️2019 CSDN 皮肤主题: 大白 设计师: CSDN官方博客