Abstract:To make the problem of multilabel classification with many classes more tractable, in recent years, academia has seen efforts devoted to performing label space dimension reduction (LSDR). Specifically, LSDR encodes high-dimensional label vectors into low-dimensional code vectors lying in a latent space, so as to train predictive models at much lower costs. With respect to the prediction, it performs classification for any unseen instance by recovering a label vector from its predicted code vector via a decoding process. In this paper, we propose a novel method, namely End-to-End Feature-aware label space Encoding (E2FE), to perform LSDR. Instead of requiring an encoding function like most previous works, E2FE directly learns a code matrix formed by code vectors of the training instances in an end-to-end manner. Another distinct property of E2FE is its feature awareness attributable to the fact that the code matrix is learned by jointly maximizing the recoverability of the label space and the predictability of the latent space. Based on the learned code matrix, E2FE further trains predictive models to map instance features into code vectors, and also learns a linear decoding matrix for efficiently recovering the label vector of any unseen instance from its predicted code vector. Theoretical analyses show that both the code matrix and the linear decoding matrix in E2FE can be efficiently learned. Moreover, similar to previous works, E2FE can be specified to learn an encoding function. And it can also be extended with kernel tricks to handle nonlinear correlations between the feature space and the latent space. Comprehensive experiments conducted on diverse benchmark data sets with many classes show consistent performance gains of E2FE over the state-of-the-art methods.

A Label Embedding Algorithm Based on Maximizing Normalized Cross-Covariance Operator

A Label Embedding Method via Conditional Covariance Maximization for Multi-label Classification.

Multilabel Feature Extraction Algorithm via Maximizing Approximated and Symmetrized Normalized Cross-Covariance Operator

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

End-to-End Feature-Aware Label Space Encoding for Multilabel Classification with Many Classes.

Cost-sensitive label embedding for multi-label classification

Label Embedding via Low-Coherence Matrices

Compact learning for multi-label classification

Maximal Correlation Embedding Network for Multilabel Learning with Missing Labels

Label Distribution Learning Through Exploring Nonnegative Components

Label Enhancement Via Joint Implicit Representation Clustering.

Maximal Linear Embedding for Dimensionality Reduction

Disentangling, Embedding and Ranking Label Cues for Multi-Label Image Recognition

Data Augmentation For Label Enhancement

Multi-View Embedding Learning for Incompletely Labeled Data.

Multi-label Classification via Feature-aware Implicit Label Space Encoding.

Multi-class Label Noise Learning via Loss Decomposition and Centroid Estimation

Incorporating Label Embedding and Feature Augmentation for Multi-Dimensional Classification.

Label Learning Method Based on Tensor Projection

An Optimized Run-Length Based Algorithm for Sparse Remote Sensing Image Labeling

MULTI-LABEL IMAGE RECOGNITION WITH JOINT CLASS-AWARE MAP DISENTANGLING AND LABEL CORRELATION EMBEDDING