Data fitting with custom model

Florian Herrla

2019 2 21

0 个回答

2 次查看（30 天）

显示更早的评论

0 个投票

Hello!

I have a measured data set of xdata and ydata (length N) where I want to determine parameters of a model. The problem is that the model includes the laminate theory of composites where for the given x (for example force) a result for y (for example strain) is calculated. Matlab only uses the full vector of the data set. If I want to get the best parameters (for example e modulus) this results in a parallel computing of N-times the laminate theory that the full vector of measured data can be used. in every help the function between x and y is mathematical like y=ax+b which is quite easy to vectorize.

Do I have to improve the code that this is possible or is there another solution for this ? An idea is to use fmincon and create a function R=(fun(parameter, xdata(1,1))-ydata(1,1))^2 and to minimize R where only one data point is used. But this result in N times the summand "(fun(parameter, xdata(1,1))-ydata(1,1))^2" and depends on the length of the given data set which is quite uncomfortable.

Are there any ideas to solve this ?