Recognition of Altered Rock Based on Improved Particle Swarm Neural Network

  • Authors:

  • Yunjun Zhan;Yanyan Wu

  • Affiliations:

  • Department of Resources and Environment Engineering, Wuhan University of Technology, Wuhan, China 430070;Department of Resources and Environment Engineering, Wuhan University of Technology, Wuhan, China 430070

  • Venue:
  • ISNN '09 Proceedings of the 6th International Symposium on Neural Networks on Advances in Neural Networks
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

PSO (particle swarm optimization) algorithm is apt to slow down and prematurity during the evolutionary anaphase. Besides, the algorithm of BP neural network also encounters some problems such as slowness in constringency, longer training time and so on. Aimed at these phenomena, PSO algorithm can be improved in two aspects: reinforcing the diversity of particles and avoiding the prematurity of particle swarm, therefore the algorithm of particle swarm neural network based on improved algorithm is presented here. Finally, this algorithm is applied to the recognition of hyper-spectral altered rock, which overcomes the disadvantage of local minimization for BP algorithm, and trained network shows great generalization ability. The instance indicates that improved PSO-BP algorithm is effective in the recognition of hyper-spectral altered rock.