Combine Bayesian regularization (trainbr) training algorithm with weight regularization.

1 回表示 (過去 30 日間)
Hi everyone.
I want to ask is it okay to train a neural network using trainbr algorithm combine with network regularization in matlab? I try to use this idea, and the result is my training always stop due to reach mu_max. If I used trainbr only the train will stop due to max epoch or max validation fail.
Here is the code that I used:
TOL=0;
net = network;
net.numInputs = 1;
net.numLayers = 6;
net.biasConnect = [1;1;1;1;1;1];
net.inputConnect = [1;1;0;0;0;0];
net.layerConnect = [0 0 0 0 0 0;
0 0 0 0 0 0;
1 1 0 0 0 0;
0 0 1 0 0 0;
0 0 1 0 0 0;
0 0 0 1 1 0];
net.outputConnect = [0 0 0 0 0 1];
net.targetConnect = [0 0 0 0 0 1];
net.inputs{1}.range = ones(NID,2);
net.inputs{1}.range(:,1) = -1;
net.layers{1}.size = NNIL;
net.layers{1}.transferFcn = TFIL1;
net.layers{1}.initFcn = 'initnw';
net.layers{2}.size = NNIL;
net.layers{2}.transferFcn = TFIL2;
net.layers{2}.initFcn = 'initnw';
net.layers{3}.size = NLPC;
net.layers{3}.transferFcn = TFIL3;
net.layers{3}.initFcn = 'initnw';
net.layers{4}.size = NNIL;
net.layers{4}.transferFcn = TFIL1;
net.layers{4}.initFcn = 'initnw';
net.layers{5}.size = NNIL;
net.layers{5}.transferFcn = TFIL2;
net.layers{5}.initFcn = 'initnw';
net.layers{6}.size = NID;
net.layers{6}.transferFcn = TFIL3;
net.layers{6}.initFcn = 'initnw';
net.initFcn = 'initlay';
net.performFcn = 'mse';
net.performParam.regularization = 0.000001;
net.trainFcn = 'trainbr';
net.trainParam.epochs = 2000;
net.trainParam.min_grad = 0;
net.trainParam.minstep = 0;
net.trainParam.max_fail = 200;
net=init(net);
net.trainParam.goal = TOL;
net.divideFcn = 'divideind';
net.divideParam.trainInd = TrainIdx;
net.divideParam.valInd = ValIdx;
net.divideParam.testInd = TestIdx;
[net,tr,Y,E,Pf,Af] = train(net,Data_All,Data_All);
Thank you very much for your attention.
  1 件のコメント
Arygianni Valentino
Arygianni Valentino 2018 年 3 月 30 日
Forget to mention, when I run this code, there is a message from matlab,
TRAINBR: NET.performParam.regularization has been set to 0.
Maybe someone can help me what does this mean?

サインインしてコメントする。

回答 (0 件)

カテゴリ

Help Center および File ExchangeNetworks についてさらに検索

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by