Hunan City University, Yiyang, Hunan, China, 413000

---

Abstract—Considering the fact that BP Neural Network has the defects of being easily falling into partial extreme value, to avoid the limitation of particle swarm algorithm we come up with an improved BP Neural Network applied in Neural Network training. This method means using nonlinear decreasing weight factor to change the fundamental ways of PSO first, and then using the developed PSO ways to optimize the original weight and threshold value of Neural Network. This method strengthens the ability of BP algorithm to solve the nonlinear problems and improve rapidity of convergence and the ability to search optimal value. We apply the improved particle swarm algorithm to agricultural land classification. Compared with BP method, this kind of algorithm can minimize errors and improve rapidity of convergence at the same time.

Index Terms—BP Neural Network, IPSO, agricultural land classification, inertia weight

[PDF]

Cite: Lianguang Mo and Zheng Xie, " An Improved BP Neural Network based on IPSO and Its Application," Journal of Computers vol. 8, no. 5, pp. 1267-1272, 2013.