mapminmax is a scaling that is applied to your input data to make it in the [-1,1] range, which you want to do. Then 'reverse' can undo the scaling, which is actually done automatically if you use your net to estimate results. If you don't apply the scaling on the inputs (which you should have a good reason not to do), you will get different results. There is a lot of information on scaling here .
'apply' 'reverse' in mapminmax
5 次查看(过去 30 天)
显示 更早的评论
hi everyone, what are 'apply' and 'reverse' in mapminmax function in neuron network? if i dont use them in mapminmax, why result in ouputs neuron network is different.
p = [0, 1, 2, 3, 4, 5, 6, 7, 8];
t = [0, 0.84, 0.91, 0.14, -0.77, -0.96, -0.28, 0.66, 0.99];
nneuron = 5;
net = feedforwardnet(nneuron);
net.inputs{1}.processFcns{2} = 'mapminmax';
net.outputs{2}.processFcns{2} = 'mapminmax';
net = configure(net, p, t);
net.trainParam.epochs = 50;
net.trainParam.goal = 0.01;
net = train(net, p, t);
y1 = sim(net, p)
pn = mapminmax('apply', p, net.inputs{1}.processSettings{2});
y2n = mySigmoidalNetwork(pn, net, nneuron);
y2 = mapminmax('reverse', y2n, net.outputs{2}.processSettings{2})
function out = mySigmoidalNetwork(in, net, nneuron) Nin = length(in);
IW = net.IW{1,1};
B1 = net.b{1};
b1 = B1(1);
LW = net.LW{2,1};
b2 = net.b{2};
h = tansig(IW*in+B1*ones(1,Nin));
out = LW*h + b2;
0 个评论
回答(1 个)
另请参阅
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!