Active Bias: Training More Accurate Neural Networks by Emphasizing High Variance Samples.
Haw-Shiuan ChangErik G. Learned-MillerAndrew McCallumPublished in: NIPS (2017)
Keyphrases
- low variance
- neural network
- training process
- training samples
- training set
- training algorithm
- variance reduction
- feedforward neural networks
- pattern recognition
- feed forward neural networks
- high precision
- bias variance decomposition
- real valued
- multi layer perceptron
- data sets
- recurrent networks
- high accuracy
- backpropagation algorithm
- back propagation
- number of training samples
- test set
- self organizing maps
- training examples
- training patterns
- training data
- wide range
- artificial neural networks
- activation function
- fuzzy logic
- highly accurate
- feed forward
- computationally efficient
- training dataset
- intra class
- semi supervised
- radial basis function network
- recurrent neural networks
- neural network model
- online learning
- machine learning
- neural network training
- bias variance
- avoid overfitting
- fault diagnosis