| |
Greedy selection algorithm by J.Friedmann
|
A=MARS(C,H) returns a mars object initialized with hyperparameters H and
based on algorithm C.
Hyperparameters, and their defaults
M = 20 -- Maximum number of base functions
d = 2 -- Trade-off in the GCV to perform model selection
C = svm -- algorithm to be used during the MARS procedure
(attribute A.child)
Model:
B -- the parameters of the base functions
Xsv -- the dataset used for learning
J -- the set of selected features (backward step of mars)
Methods:
training, testing
Example:
[r a]=train(mars(svm('ridge=0.001'),'M=2'),toy2d('l=30'))
| |
Reference : Multivariate adaptive regression splines |