Partial Label Learning via Cost-Guided Retraining, 27TH EUROPEAN CONFERENCE ON ARTIFICIAL INTELLIGENCE(ECAI 2024)

Address the label imbalance in PLL via cost sensitive.

Matlab Implementation

1. How to start?

   To help you start working with PL-CGR, please run demo.m.

2. Parameter set

   mu = 0.04, lambda = 0.6, c = 0.2: lost, MSRCv2

   mu = 0.05, lambda = 0.4, c = 0.2: Mirflickr, malagasy, Italian

   mu = 0.05, lambda = 0.5, c = 0.2: FG-NET

   mu = 0.05, lambda = 0.6, c = 0.2: Yahoo! News, Soccer Player