What can be the possible benefits of employing Neural Networks rather than Genetic Programming and its less complicated linear variants such as GEP and MEP?
You may follow the below links.....
https://pdfs.semanticscholar.org/f82e/665647e24d9b8dd50df9132c5240c1b96d6b.pdf
http://www.engineeringletters.com/issues_v14/issue_1/EL_14_1_18.pdf
https://link.springer.com/article/10.1007/s11269-011-9852-8
https://www.researchgate.net/profile/Robert_Dorsey/publication/222477189_Toward_global_optimization_of_neural_networks_A_comparison_of_the_genetic_algorithm_and_backpropagation/links/0c960516e984aa7b54000000.pdf
http://ira.lib.polyu.edu.hk/bitstream/10397/1133/1/IJEP3.pdf
Article Toward global optimization of neural networks: A comparison ...
NN is a way to approximate arbitrary functions, with applications in classification and regression.
Genetic algorithms (GA) are optimization techniques, which can be used to learn NNs.
I dont think the question is whether to use NN or GA, but how GA can be used for better training of NNs
Dear Dr. Najla,
For NN:
Advantages are:
1- don't required a powerful computers
2- support multi-inputs and multi-outputs
3- very successful in recognizing patterns, classifications and control applications
Disadvantages are:
1- its product is a "black box", net structure and weight matrix are embedded in the used software.
For GP:
Advantages are:
1- The product is closed form mathematical formula, it could be used manually of by software.
2- it is very successful in multi-variable regression applications
3- it found the optimum structural for the mathematical formula without used help
Disadvantages are:
1- it required a powerful computer ( in learning phase )
2- can't be used in run-time applications like control
3- support multi-inputs put only single output
I hope this helps you
regards
- Badges
- Science topic