A heuristically enhanced gradient approximation (HEGA) algorithm for training neural networks.
Dimokritos PanagiotopoulosChristos OrovasDimitrios SyndoukasPublished in: Neurocomputing (2010)
Keyphrases
- neural network
- learning algorithm
- dynamic programming
- neural network training
- recognition algorithm
- preprocessing
- search space
- cost function
- training process
- error bounds
- detection algorithm
- segmentation algorithm
- simulated annealing
- optimal solution
- optimization algorithm
- feed forward
- approximation algorithms
- similarity measure
- np hard
- training phase
- feed forward neural networks
- input pattern
- approximation ratio
- decision trees