Push-pull separability objective for supervised layer-wise training of neural networks.
Lech SzymanskiBrendan McCanePublished in: IJCNN (2012)
Keyphrases
- neural network
- training process
- multi layer
- training algorithm
- supervised learning
- feed forward neural networks
- error back propagation
- feedforward neural networks
- neural network training
- back propagation
- artificial neural networks
- pattern recognition
- training set
- backpropagation algorithm
- output layer
- hidden layer
- supervised training
- restricted boltzmann machine
- test set
- supervised methods
- training patterns
- pairwise
- multi layer perceptron
- feed forward
- genetic algorithm
- training stage
- online learning
- dimensionality reduction
- learning algorithm
- network architecture
- semi supervised
- single layer
- data sets
- recurrent networks
- activation function
- neural nets
- multilayer perceptron
- conditional random fields
- self organizing maps
- training examples
- unsupervised learning
- hidden markov models