SEMKL.regression: Prediction from MKL model

In cwilso6/RMKL: Multiple Kernel Learning for Classification, Regression Problems, and Survival Settings

Description

This makes prediction for multiple kernel regression

Usage

SEMKL.regression(k, outcome, penalty, epsilon, tol = 1e-04, max.iters = 1000)

Arguments

k	Gramm matrix of training data
outcome	observed dependent variable
penalty	Cost of unit miss fitted - observed
epsilon	SVM parameter defining support vectors
tol	Convergence criteria, algorithm stops once the biggest change of kernel weights in two consectutive iterations is less than tol.
max.iters	Termination criteria, algorithm will stop after 1000 iterations

Value

results Returns a list which includes model parameters, weights for kernels, and f which are the fitted values for the training set

Examples

## Not run: 
library(kernlab)
x=as.matrix(10*runif(200),ncol=1)
y=x*sin(x)+rnorm(200)
plot(x,y)


model=SEMKL.regression(k=K$K.train, outcome = y[1:150], epsilon = 0.25, penalty=100)

## End(Not run)

cwilso6/RMKL documentation built on May 18, 2021, 9:58 a.m.