Combining stochastic average gradient and Hessian-free optimization for sequence training of deep neural networks.
Pierre L. DogninVaibhava GoelPublished in: ASRU (2013)
Keyphrases
- neural network
- training process
- training algorithm
- stochastic optimization
- feedforward neural networks
- feed forward neural networks
- pattern recognition
- stochastic programming
- multi layer perceptron
- optimization algorithm
- stochastic search
- optimization problems
- edge detection
- deep architectures
- training phase
- neural network structure
- global optimization
- monte carlo
- error back propagation
- backpropagation algorithm
- stochastic context free grammars
- multi layer
- neural nets
- highly non linear
- back propagation
- supervised learning
- weighted sums
- steepest descent method
- optimization process
- standard deviation
- multilayer perceptron
- test set
- training examples
- evolutionary algorithm
- objective function