Abstract:Dimensionality reduction is a key technique in the data mining field.Conventional dimensionality reduction methods are used only for single label learning.Although supervised dimensionality reduction has been studied widely,little work on multi-label dimensionality reduction has been done due to the complexity of multi-label learning.Direct application of existing unsupervised dimensionality reduction methods to multi-label tasks ignores the label information.We can deploy single label dimensionality reduction methods to solving multi-label problems by decomposing these problems into single label ones.However,this kind of method does not consider the correlations between the different labels of each instance.The only method for multi-label dimensionality reduction is MDDM(multi-label dimensionality reduction via dependency maximization),which is a feature extraction method.Multi-label feature selection remains almost untouched for the complexity of evaluating the goodness of features in multi-label learning.Therefore,an embedded method of feature selection named MEFS(multi-label embedded feature selection) is proposed,which employs the prediction risk criterion as evaluation metric for features.We perform two experiments on Yahoo web page categorization data sets which are widely used for benchmark evaluation:(1) analyzing the effectiveness of multi-label evaluation metrics on evaluating features;(2) comparing MEFS and MDDM,PCA(principal component analysis),LPP(locality preserving projections component analysis),and experiments show that the performance of MEFS is superior to that of some state-of-arts multi-label dimensionality reduction methods.

Multi-label Local Discriminative Embedding

Multi-label Linear Discriminant Analysis with Locality Consistency.

Two-Stage Label Embedding Via Neural Factorization Machine for Multi-Label Classification

Multi-label Learning with Discriminative Features for Each Label

Dependence Maximization Based Label Space Dimension Reduction for Multi-Label Classification.

Multi-Label Low-dimensional Embedding with Missing Labels

Semi-Supervised Multi-Label Linear Discriminant Analysis

Multi-dimensional Multi-Label Classification: Towards Encompassing Heterogeneous Label Spaces and Multi-Label Annotations

Embedded feature selection for multi-label learning

Semi-Supervised Multi-Label Dimensionality Reduction Based on Dependence Maximization.

Multi-Label Learning by Exploiting Label Correlations Locally

Coupled Dictionary Learning for Multi-label Embedding.

Multi-Directional Multi-Label Learning

Multi-label Feature Selection with Local Discriminant Model and Label Correlations.

Multi-Label Learning With Local Similarity Of Samples

Exploring Instance Correlations with Local Discriminant Model for Multi-Label Feature Selection

Multi-Label Learning by Exploiting Label Correlations with LDA

A Weighted Linear Discriminant Analysis Framework for Multi-Label Feature Extraction

Multi-Dimensional Classification Via Class Space Fusion and Comprehensive Label Correlations

Semi-Supervised Multi-Label Dimensionality Reduction Learning by Instance and Label Correlations

Multilabel Dimensionality Reduction Via Dependence Maximization.