Prev: TriScatteredInterp
Next: Convolution of Z-transform 2D filter kernel with 2D matrix
From: Kyle on 19 May 2010 11:33
I have two to five explanatory variables for a linear regression model. Each explanatory variable has a hyperparameter (distance). Thus each explanatory variable changes slightly when the hyperparameter is adjusted. I want to find the combination of explanatory variables that maximizes the r-squared in the model. I have the explanatory variables calculated at each hyperparameter increment which is up to 200 different hyperparameter values.
I've done a brute force method of finding the best model for a bivariate case, but if I want to do three or more variables the computation time gets huge, so an optimization approach here seems applicable.
I need help on setting up this optimization problem. Thanks in Advance.
Kyle
 | 
Pages: 1
Prev: TriScatteredInterp
Next: Convolution of Z-transform 2D filter kernel with 2D matrix