Training multi-layer perceptron (MLP) is non-trivial due to its non-linear nature and the presence of large number of local optima. Meta-heuristic algorithms may solve this problem efficiently. In this paper, we investigate the use of glowworm swarm optimisation (GSO) algorithm in training the MLP neural network. The GSO based trainer is evaluated on five classification datasets, namely Wisconsin breast cancer, BUPA liver disorders, vertebral column, exclusive OR (XOR) and balloons. The evaluations are conducted by comparing the proposed trainer with four other meta-heuristics, namely biogeography-based optimisation (BBO), genetic algorithm (GA), bat (BAT) and multi-verse optimiser (MVO) algorithms. The results show that our proposed trainer achieves better classification accuracy rate in most datasets compared to the other algorithms.
|Title of host publication||Proceedings of the Fourth IEEE/ACM International Conference on Big Data Computing, Applications and Technologies (BDCAT '17)|
|Number of pages||8|
|Publication status||Published - 31 Dec 2017|
- multi-layer perceptrons
- neural network
- glowworm swarm optimisation (GSO)
Alboaneen, D. A., Tianfield, H., & Zhang, Y. (2017). Glowworm swarm optimisation for training multi-layer perceptrons. In Proceedings of the Fourth IEEE/ACM International Conference on Big Data Computing, Applications and Technologies (BDCAT '17) (pp. 131-138 ). IEEE. https://doi.org/10.1145/3148055.3148075