Using If statement to train Neural Network

Question

0 个投票

Good day, Is it possible to use if statement to train a neural network ? I have specific set of inputs that correspond with specific set of outputs with different sizes. For example, the 3 set of inputs are a 24x2 matrix each corresponding with 3 sets of output 5x2,7x2 and 9x2 respectively.

After training, the simulation only gives me back for 5 outputs with very bad results. But I need it give me back for 5, 7 and 9.

I don't know if it's my input argument (eg. ' if network2.layers{2}.size == 7 & network2.inputs{1}.exampleInput==i2') that's the issue or I'm going about it the wrong way. Please I need help.

Thank you for any input response.

Below is the code written. CODE network2=network(1,2,[1;1],[1;0],[0 0;1 0],[0 1]); network2.inputs{1}.size=24; network2.inputs{1}.processFcns = {'removeconstantrows','mapminmax'}; network2.layers{1}.size = 30; network2.layers{1}.transferFcn='tansig'; network2.layers{1}.initFcn = 'initnw'; network2.layers{2}.size = 5; network2.layers{2}.transferFcn='purelin'; network2.layers{2}.initFcn = 'initnw'; network2.inputs{1}.exampleInput=i1; network2.initFcn = 'initlay'; network2.performFcn = 'mse'; network2.trainFcn = 'trainlm'; network2.divideFcn = 'dividerand'; network2.plotFcns = {'plotperform','plottrainstate','plotregression'}; network2.trainParam.epochs=10000; network2.trainParam.goal=0; network2.trainParam.min_grad=1e-7; if network2.layers{2}.size == 5 & network2.inputs{1}.exampleInput==i1 train(network2,i1,o1); else if network2.layers{2}.size == 7 & network2.inputs{1}.exampleInput==i2 train(network2,i2,o2); else train(network2,i3,o3); end end DATASET i1=[281.907400000000,1198.04650000000;284.647200000000,1201.21840000000;236.376200000000,1191.29540000000;183.152000000000,1174.05820000000;135.078500000000,1144.08000000000;95.9512000000000,1086.29350000000;72.3133000000000,983.455900000000;61.6324000000000,793.240500000000;62.4501000000000,562.124400000000;69.2182000000000,352.981300000000;85.8570000000000,172.069000000000;109.075800000000,110.127400000000;140.669700000000,103.392200000000;178.568700000000,118.533800000000;224.686200000000,140.450000000000;289.353400000000,166.471100000000;361.089100000000,191.471700000000;436.180400000000,213.347100000000;540.414000000000,237.769600000000;639.630600000000,255.883300000000;740.965800000000,270.163300000000;833.914200000000,280.210200000000;918.818500000000,287.279300000000;1002.77820000000,292.616800000000] i2=[20.0326000000000,59.6496000000000;19.9899000000000,60.0313000000000;20.1479000000000,58.4751000000000;20.4259000000000,55.9220000000000;20.9007000000000,51.9693000000000;21.7638000000000,45.7339000000000;23.1983000000000,37.2975000000000;25.5834000000000,26.7863000000000;28.2221000000000,19.1738000000000;30.5521000000000,15.5267000000000;33.2405000000000,14.9470000000000;36.3009000000000,17.0807000000000;41.6112000000000,20.9500000000000;50.3339000000000,26.1645000000000;63.3688000000000,33.1481000000000;84.3511000000000,43.9976000000000;110.013400000000,57.3753000000000;139.548200000000,72.8986000000000;185.873800000000,97.4243000000000;237.205600000000,124.815400000000;299.394400000000,158.303300000000;368.616500000000,195.977800000000;446.956500000000,239.142900000000;547.245400000000,295.276400000000] i3=[10.2122000000000,89.8217000000000;10.1287000000000,90.6532000000000;10.9130000000000,106.931400000000;12.1459000000000,130.142200000000;13.8731000000000,158.514900000000;16.3103000000000,193.140200000000;19.4212000000000,232.965300000000;23.6388000000000,281.420900000000;28.0371000000000,325.456600000000;32.3947000000000,360.688000000000;38.2749000000000,393.106300000000;43.9239000000000,410.107900000000;49.8511000000000,420.214700000000;55.3723000000000,432.560300000000;60.4935000000000,461.510300000000;66.5390000000000,532.130800000000;74.1069000000000,643.727300000000;84.8775000000000,781.726000000000;104.945700000000,992.866000000000;128.719100000000,1209.10510000000;156.821300000000,1446.89610000000;186.298700000000,1683.43430000000;217.292600000000,1919.24040000000;253.457200000000,2177.39820000000] o1=[300,1200;50,50;1200,300;1.30000000000000,5;10,12] o2=[20,60;60,10;22,40;1200,700;4.10000000000000,3;3.60000000000000,10.8000000000000;12.5000000000000,19.5000000000000] o3=[10,84;50,730;100,400;20,280;400,3000;2,1.40000000000000;11,4;20,15;15,20]