Label Embedding via Low-Coherence Matrices

Jianxin Zhang, Clayton Scott

Label embedding is a framework for multiclass classification problems where each label is represented by a distinct vector of some fixed dimension, and training involves matching m…