Evolutionary Algorithms for Neural Network Weight Optimisation

This repo contains the code for neural network weight optimisation using 4 evolutionary algorithms, namely:

• Genetic Algorithm
• Cultural Algorithm
• Ant Colony Optimisation
• Particle Swarm Optimisation

ANN Architecture

Layer	Number of Neurons	Activation Function
Hidden	12	ReLU
Hidden	8	ReLU
Hidden	16	ReLU
Hidden	8	ReLU
Output	2	Softmax

Params for Ant Colony Optimization

Generations	Ants	Decay Constant	Training Accuracy	Testing Accuracy
10	5	0.1	93.66%	94.37%

Params for Cultural Algorithm

Generations	Population Size	Training Accuracy	Testing Accuracy
3	10	93%	94%

Params for Genetic Algorithm

Generations	Population Size	Parent Selection	Number of Parents	Training Accuracy	Testing Accuracy
20	5	Roulette Wheel	2	93.66%	94.37%

Params for Particle Swarm Optimisation

Generations	Population Size	c1	c2	Inertial Weight	Fitness of Best Particle	Testing Accuracy
10	10	2	2	0.8	4.234	94%

References:

Genetic Algorithm

Montana, David J., and Lawrence Davis. "Training feedforward neural networks using genetic algorithms." IJCAI. Vol. 89. 1989.

Cultural Algorithm

Reynolds, Robert G. "An introduction to cultural algorithms." Proceedings of the 3rd annual conference on evolutionary programming, World Scientific Publishing. 1994.

Ant Colony Optimisation

Mavrovouniotis, Michalis, and Shengxiang Yang. "Evolving neural networks using ant colony optimization with pheromone trail limits." 2013 13th UK Workshop on Computational Intelligence (UKCI). IEEE, 2013.

Particle Swarm Optimisation

Mazaheri, Pooria, et al. ‘Designing Artificial Neural Network Using Particle Swarm Optimization: A Survey’. Swarm Intelligence - Recent Advances and Current Applications, IntechOpen, 8 Feb. 2023. Crossref, doi:10.5772/intechopen.106139.