weightdecayaddnoise.m

有监督自组织映射-偏最小二乘算法(A supervised self-organising map–partial least squares algorithm),可以用语多变量数据的回归分析

function [DecMap] = WeightDecayAddNoise(Map, DecayFract, MLKP);

DecMap=Map;
if (upper(MLKP.AddNoise) == 'Y')
    [Nvar,Nunits]=size(Map);
    MaxVal=max(Map');
    MinVal=min(Map');
    RangeFract=(MaxVal-MinVal)*MLKP.NoiseFraction;
    Noise=(rand(Nvar,Nunits)-0.5)*(1-DecayFract);
    for ivar=1:Nvar
        Noise(ivar,:) = Noise(ivar,:)*RangeFract(ivar);
    end
    DecMap=DecMap+Noise;
end

if (upper(MLKP.WeightDecay) == 'Y')
    DecMap=DecMap*DecayFract;
end