基于改进粒子群算法的BP算法的研究6.pdfVIP

28 2 2011 2 : 1006 - 9348( 20 11) 02 - 0 147- 04 BP 张 丹, 韩胜菊, 李 建, 聂尚宇 ( , 116023) : BP BP , , BP , , , BP BP , , BP BP , BP : ; ; : TP 30 16 : A BPAlgorithm Basedon ImprovedParticle Swarm Optimization ZHANG Dan, HAN Sheng- u, L I J ian, N IE Sh ang- yu ( Co llege of Inform a tion Eng ineering, D alian F isher ies U n iverc ity, D a lian L iaon ing 116023, Ch ina) ABSTRACT: In th is paper, P roposed BP a lgo rithm based on mi proved PSO algor ithm w hich adaptive dynam ic ad u stment o f ine rtia w eight, by ana lyzing the shortcom ings o f BP a lgo rithm and the de fic iency o f u sing standa rd pa rti cles sw arm optmi ization BP ne wt ork w e ights. The ine rtia w e ights w ere ad usted acco rd ing to the chang es o f f itness, and not depended on the m ax mi um num ber o f itera tions and contem po ra ry iterations, so that the entire netw o rk has a faster convergence speed and sm a ller errors. T hen, it w as u sed in the d iagno sis o f d iseases o f sea cucumbers. Itw as found that the BP algor ithm based on adaptive dynam ic ad ustment of inertia w e ight PSO a lgor ithm has faste r conver gence than the BP a lgor ithm based on PSO algor ithm and has higher accuracy rate, and the BP netw o rks based on the mi proved algor ithm is stable than the BP ne wt orks based on PSO a lgo rithm. T his paper argues that u sing the PSO al go rithm w hich adaptive dynam ic ad u stm ent o f ine rtia w e ight to optmi iza tion BP a lgor ithm is be tter than u se PSO a lgo rithm optmi iza tion BP a lgor ithm. KEY ORDS: Partic le sw arm optmi ization( PSO ) ; Ine rtia w e igh; F itness