Suppose we have N points in the k-dimensional space [i.e. each a k-tuple of the form x=(x1,x2,...,xk)]. We need to find the nonlinear least-square surface F (ex. polynomial) which minimizes the sum of the residuals ||x-F||. In theory this is a simple optimization problem but requires tedious implementations. What are the best software packages in this regard?
Suggestions about nonlinear regression methods are also welcome.
JMP can analyse your data within few minute if you know the usage.
If you have access to Matlab, you can use it's tools like lsqnonlin or fminsearch. I hope you can figure out how to implement the cost function.
Thank you everyone. I went ahead with the matlab function fitnlm though it does not address the most general case. The surface in this case is x_k=F(x_1,x_2,...,x_{k-1}) instead of G(x_1,...x_k)=0. But it is serving my purpose (for now at least).
I think the R package "nlme" and MATLAB functions "fmincon" and "fminunc" are useful ways.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software