pragmahdrstopi
cludestdiohi
cludeiostreamhco
stA300co
stB100co
stMAX500最大训练次数co
stCOEF00035网络的学习效率co
stBCOEF0001网络的阀值调整效率co
stERROR0002网络训练中的允许误差co
stACCURACY00005网络要求精度doublesample4140000514190205331415055214360533141505321539053215390551196805122106053314150554126805521436051319610534136505551243051419020514190205351339055412680513196105321539013111110152652111310190134604315552421535724114976613558701545406113101901159545134604315357241121125013111110133638015265211111600013272191535724doublew41010wc41010b410bc410doubleo410
eti
410d410differ单个样本的误差doubleis全体样本均方差i
tcou
tavoid
etouti
tmi
t
计算网络隐含层和输出层的输出voidcalculdi
tmi
t
计算网络的反向传播误差voidcalcalwci
tmi
t
计算网络权值的调整量voidcalcaulbci
tmi
t
计算网络阀值的调整量voidcha
gewi
tmi
t
调整网络权值voidcha
gebi
tmi
t
调整网络阀值voidclearwci
tmi
t
清除网络权值变化量wcvoidclearbci
tmi
t
清除网络阀值变化量bcvoidi
itialwvoid初始化NN网络权值Wvoidi
itialbvoid初始化NN网络阀值voidcalculdiffervoid计算NN网络单个样本误差voidcalculisvoid计算NN网络全体样本误差voidtrai
NNvoid训练NN网络计算NN网络隐含层和输出层的输出void
etouti
tmi
t

fi
tijk隐含层各节点的的输出forj1i2jmjm为隐含层节点个数
eti
ij00fork1k3k隐含层的每个节点均有三个输入变量
eti
ij
eti
ijoi1kwikj
eti
ij
eti
ijbijoijA1exp
eti
ijB输出层各节点的输出forj1i3j
j
eti
ij00fork1kmk
eti
ij
eti
ijoi1kwikj
eti
ij
eti
ijbijoijA1exp
eti
ijB计算NN网络的反向传播误差voidcalculdi
tmi
t
i
tijkdoubletacou
t1d31o31samplea3ABexp
eti
31Bpow1exp
eti
31B2隐含层的误差forj1i2jmjt000fork1k
kttwi1jkdi1kdijtABexp
eti
ijBpow1exp
eti
ijB2计算网络权值W的调整量voidcalculwci
tmi
t
i
tijk输出层（第三层）与隐含层（第二层）之间的连接权值的调整fori1k3imiforj1j
j
fwcr