6.3 Synaptic convergence to centroids: AVQ Algorithms Unsupervised Competitive Learning (UCl m(t+1)=m,(m)+c1[x(t)-m(t) m(+1)=m1(t)fi≠j ic, defines a slowly decreasing sequence of learning coefficient For instance, C,=0.11 for 10,000 samples x(t) 10.000 Supervised Competitive Learning(sCl m(+1)=m()+cr,(xO)x()-m() m, (t)+cIx(t-m t if XED m, (t)-c[(t-m, o if xE D 2003.11.19
2003.11.19 11 6.3 Synaptic convergence to centroids: AVQ Algorithms Unsupervised Competitive Learning (UCL) ( 1) ( ) [ ( ) ( )] ( 1) ( ) j j j t i i m t m t c x t m t m t m t if i j + = + − + = { }t c defines a slowly decreasing sequence of learning coefficient For instance , 0.1 1 for 10,000 samples ( ) 10,000 t t c x t = − Supervised Competitive Learning (SCL) ( 1) ( ) ( ( )) ( ) ( ) ( ) [ ( ) ( )] ( ) [ ( ) ( )] j j t j j j t j j j t j j m t m t c r x t x t m t m t c x t m t if x D m t c x t m t if x D + = + − + − = − −
6.3 Synaptic convergence to centroids: AvQ algorithms Differential Competitive Learning dcl m(t+1)=m(1)+cAS(y,(m)x()-m,() m(t+1)=m2(t)yfi≠j A,(,() denotes the time change of the ith neuron's competitive signal S,(y) AS(y()=S(y/(t+1)-S(y() practice we often use only the sign of the signal difference or sgn(4y I the sign of the activation difference 2003.11.19 12
2003.11.19 12 6.3 Synaptic convergence to centroids: AVQ Algorithms ( ) S y j j sgn[ ] j y Differential Competitive Learning (DCL) ( 1) ( ) ( ( ))[ ( ) ( )] ( 1) ( ) j j t j j j i i m t m t c S y t x t m t m t m t if i j + = + − + = ( ( )) j j S y t denotes the time change of the jth neuron’s competitive signal ( ( )) ( ( 1)) ( ( )) = + − S y t S y t S y t j j j j j j In practice we often use only the sign of the signal difference or , the sign of the activation difference
=C0 影响迭代步长 T m(t+)=m()+Cx(0)-m() 终止迭代条件:m()m(+<E 总迭代次数 样本数 总迭代次数可以 样本数决定计算 人为设定? 时间及精度 计算时间及精度终止迭代的条件 可以人为设定? 是否不需要? 2003.11.19 13
2003.11.19 13 总迭代次数 样本数 样本数决定计算 时间及精度 总迭代次数可以 人为设定？ 终止迭代的条件 是否不需要？ 计算时间及精度 可以人为设定？ T 0 1 ( 1) ( ) [ ( ) ( )] ( ) ( 1) t j j t j j j k c c T m t m t c x t m t m t m t = − + = + − + 影响迭代步长 终止迭代条件： －