SEMKL.classification: Simple and Efficient MKL

In RMKL: Multiple Kernel Learning for Classification or Regression Problems

Description

This function conducts Simple and Efficnent MKL for precomputed gramm matrices

Usage

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

Arguments

k	list of Gramm matrices
outcome	vector of binary outcome -1 and 1
penalty	penalty of the smoothness of the resulting desicion rules
tol	change between to iterations is smaller than this, algorithms is considered to have converged
max.iters	maximum number of allowed iteratons

Value

gamma weight vector for the importnace of each kernel

alpha coeffiencents of the dual of MKL

time total amount of time to train model

iters Numvber of iterations to reach convergence criteria

gamma_all Kernel weights for each interation of SEMKL

Examples

data(benchmark.data)
example.data=benchmark.data[[1]]
#Load data
training.samples=sample(1:dim(example.data)[1],floor(0.7*dim(example.data)[1]),replace=FALSE)
# Split samples into training and test sets 
C=1
kernels=c('radial','polynomial')
degree=c(0,2)
scale=c(0,2)
sigma=c(2,0)
K=kernels.gen(example.data[,1:2], training.samples, kernels, degree, scale, sigma)
K.train=K$K.train
SEMKL.classification(K.train,example.data[training.samples,3], C)

RMKL documentation built on May 2, 2019, 7:55 a.m.