清空筛选条件
清空筛选条件

Help in configuration of backpropagation neural network LM

1 次查看（过去 30 天）
Rafa
Rafa 2011-11-1
I am applying to a prediction problem, where the type of network is supervised, the backpropagation algorithm and Levenberg Marquadt. At least so far.
I am using the command newff and i am using MatLab2008a, Neural toolbox.
The input matrix is 3x125 with 3 input variables. The array of output or target is 3 x 125, representing three variables'. 125 samples are used for training.
Thank you So far I have one hidden layer set up but I intend to find a satisfactory performance.
I'm using as activation function for hidden layer and output type logsig.
Questions: 1-one could give me tips on adjusting the parameters: net.trainParam.show = 5;% Updates the display net.trainParam.epochs = 30000;% Maximum epochs net.trainParam.goal = 0005;% error rate (error Goal) net.trainParam.lr = 0.0005;% Rate of learning net.trainParam.mc = 0.98
2-The range of the input array ranges from 0 to 20 and the output matrix varies from 10 to 6,800,000. I received a reply a while ago just to use escala.Mas reduce both output and input?
3-Levenberg Marquadt is right?
4-The problem is an application of a control system in a typical non-linear situation.

请先登录，再回答此问题。

回答（1 个）

Greg Heath
Greg Heath 2011-12-1
I see one glaring problem. However, just use all of the default values. If results are unsatisfactory, then think about changing them.
Post relevant code if you are still having problems.
Hope this helps.
Greg

类别

在 Help Center 和 File Exchange 中查找有关 Image Data Workflows 的更多信息

标签

Community Treasure Hunt

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

Start Hunting!

Translated by