Title :
A global optimization algorithm for neural network training
Author_Institution :
Dept. of Electr. & Comput. Syst. Eng., Monash Univ., Clayton, Vic., Australia
Abstract :
The main thrust of the research is to develop a global optimization algorithm tailored for multilayer feedforward back-propagation neural network training. The goal in designing the algorithm is to tackle the problem of reaching nonoptimal network configurations due to being trapped by a saddle point or a local minimum so that continuous learning through automatic online retraining is feasible.
Keywords :
backpropagation; feedforward neural nets; multilayer perceptrons; optimisation; automatic online retraining; continuous learning; global optimization algorithm; local minimum; multilayer feedforward back-propagation neural network training; nonoptimal network configurations; saddle point; Algorithm design and analysis; Australia; Computer networks; Design optimization; Feedforward neural networks; Humans; Multi-layer neural network; Neural networks; Neurons; Systems engineering and theory;
Conference_Titel :
Neural Networks, 1993. IJCNN '93-Nagoya. Proceedings of 1993 International Joint Conference on
Print_ISBN :
0-7803-1421-2
DOI :
10.1109/IJCNN.1993.713950
