SIGUA: Forgetting May Make Learning with Noisy Labels More Robust.
Bo HanGang NiuXingrui YuQuanming YaoMiao XuIvor W. TsangMasashi SugiyamaPublished in: ICML (2020)
Keyphrases
- learning process
- incremental learning
- online learning
- supervised learning
- learning algorithm
- learning systems
- computationally efficient
- prior knowledge
- reinforcement learning
- pairwise
- multi agent
- image sequences
- knowledge acquisition
- unsupervised learning
- case study
- social networks
- learning scenarios
- genetic algorithm
- label noise