Euclidean matrix distance for radial basis neural network

[altenbach · Accepted solution]

Yes, many things could be simplified. You don't need matrices or any matrix operations.

(I had to add some formatting code because your decimal delimiter is comma, which the US version does not understand directly, thus the "%,;" in the format)

Modify as needed. As I said most of your file data is 1D, so don't drag it along as 2D arrays.

Note that you could substitute the dot product for the last Multiply&Sum but it probably does not make a difference, except that it requires loading lvanly.dll, which is otherwise not needed. It is also a bit more picky, for example it will fail if the two arrays differ in length while the above code just let the shorter size win.)

Not extensively tested. Please verify correct operation.

[Kamila_Jankowska]

It works good and looks pretty smart, thank you very much. I am going to try optimize perceptron and Kohonen map either. In fact I used matrixes because I thought it will be good for MLP network but in here it was a source of trouble.

[altenbach]

It has been fun to study this. Good luck with your project. 🙂