Problem with mathematical expression of neural network

Question

Sree Ranjini 2018-4-12

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/394478-problem-with-mathematical-expression-of-neural-network

回答： Ilias 2018-4-21

Hello,

I am using 'feedforwardnet' to model ANN. After training ANN, I use 'getNeuralNetExpression()' to obtain symbolic expression of neural network. When this expression is applied on set of input values in MS Excel, I get a wrong output. But when I use net() function on input values in MATLAB, I get output close to the target values. What may be the reason for this? Input values are not normalized to (0 1) range. Is this problem related to normalization of inputs?

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Greg Heath 2018-4-13

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/394478-problem-with-mathematical-expression-of-neural-network#answer_314898

You can answer that question better than we can.

Normalize your data and see what happens.

Or am I missing something ???

Greg

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

Answer 2

Ilias 2018-4-21

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/394478-problem-with-mathematical-expression-of-neural-network#answer_316434

If you haven't changed the default setting, inputs are normalized at [-1 1]. https://se.mathworks.com/help/nnet/ug/choose-neural-network-input-output-processing-functions.html

You can use the genFunction instead: https://se.mathworks.com/help/nnet/ref/genfunction.html