Parameter estimation from a user specific algorithm
1 回表示 (過去 30 日間)
古いコメントを表示
I have a complex physical system which I can model through an algorithm which I've written in Matlab. It's not a standard state-space, ODE, etc. model, it's an algorithm for which I specify input parameters and I obtain an output which I can match to data which I take from the real physical system.
Is there any tool in Matlab which allows me to run my algorithm with different parameter values and try to estimate the parameters for a best fit to my data? I've tried writing an MCMC parameter walk with least squares fitting, but its a work in progress and quite inefficient at the moment, and was wondering if there was something out there already that could do this for me.
Thanks!
1 件のコメント
Muthu Annamalai
2013 年 2 月 27 日
Do you mean Monte Carlo technique by MCMC ? I think you can try to use rand() in a for-loop to pickout the various parameters and find fitness functions to see goodness of fit from your solver/curve-fit solutions.
You can try to see if there is a match with MATLAB System ID toolbox http://www.mathworks.com/products/sysid/examples.html?s_cid=BB or the MATLAB genetic algorithm http://www.mathworks.com/help/gads/genetic-algorithm.html
回答 (1 件)
Matt J
2013 年 2 月 26 日
Basically any tool in the Optimization Toolbox could be applicable, depending on the constraints (if any) you're imposing on the parameters. Look at LSQNONLIN or FMINCON.
0 件のコメント
参考
カテゴリ
Help Center および File Exchange で Get Started with Curve Fitting Toolbox についてさらに検索
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!