Active Bias: Training a More Accurate Neural Network by Emphasizing High Variance Samples.
Haw-Shiuan ChangErik G. Learned-MillerAndrew McCallumPublished in: CoRR (2017)
Keyphrases
- neural network
- low variance
- training algorithm
- training samples
- training set
- training process
- feed forward neural networks
- training patterns
- high precision
- neural network training
- variance reduction
- feedforward neural networks
- multi layer perceptron
- wide range
- training examples
- back propagation
- pattern recognition
- number of training samples
- highly accurate
- real valued
- avoid overfitting
- artificial neural networks
- train a neural network
- bias variance decomposition
- data sets
- recurrent neural networks
- genetic algorithm
- training data
- feed forward
- self organizing maps
- high accuracy
- knn
- training dataset
- bias variance
- neural network model
- backpropagation algorithm
- sample points
- machine learning
- high quality
- trade off
- test set
- standard deviation
- training phase
- correlation coefficient
- multilayer perceptron
- associative memory