Code for our TNNLS'21 paper titled Label Distribution Learning by Exploiting Label Distribution Manifold
Our paper presents a new LDL method which exploits both global and local label correlations. It uses the manifold strucutre of label distribution to model label correaltion, which doesn't rely on any assumption and is totally data-dependent.
Use our code and cite
@article{wang_label_2021,
title = {Label distribution learning by exploiting label distribution manifold},
journal = {IEEE Transactions on Neural Networks and Learning Systems},
author = {Wang, Jing and Geng, Xin},
year = {2021},
pages = {1--14},
}
LDL-LDM with full LDL: python ldm_full.py
LDL-LDM with missing LDL: python ldm_incom.py
IIS-LLD and SA-BFGS: http://ldl.herokuapp.com/LDLPackage_v1.2.zip
Adam-LDL-SCL, EDL-LRL, and LDL-LCLR, refer to https://github.com/NJUST-IDAM/