Pumpout: A Meta Approach for Robustly Training Deep Neural Networks with Noisy Labels.
Bo HanGang NiuJiangchao YaoXingrui YuMiao XuIvor W. TsangMasashi SugiyamaPublished in: CoRR (2018)
Keyphrases
- neural network
- training process
- training algorithm
- training set
- training examples
- feed forward neural networks
- multi layer perceptron
- training data
- pattern recognition
- training patterns
- feedforward neural networks
- backpropagation algorithm
- deep architectures
- back propagation
- training phase
- neural network training
- test set
- artificial neural networks
- genetic algorithm
- noisy environments
- ground truth labels
- recurrent neural networks
- neural network model
- pairwise
- recurrent networks
- machine learning
- annotation effort
- multilayer neural network
- error back propagation
- e learning
- labelled data
- fuzzy logic
- low quality
- supervised learning
- deep learning
- missing data
- feed forward
- neural nets
- fuzzy neural network