Description Usage Arguments Value Examples
This function conducts DALMKL for precomputed gramm matrices
1 2 3 |
K |
The multiple kernel cube (3-d array) |
y |
The outome variable, must be -1/1 |
loss |
The loss function to be used, must be either 'hinge' or 'logistic', default to be 'hinge' |
C |
tuning parameter for block one norm, default to be .5 |
tolOuter |
change between to iterations is smaller than this, algorithms is considered to have converged for outer loop, default to be .01 |
tolInner |
change between to iterations is smaller than this, algorithms is considered to have converged for inner loop, default to be .000001 |
OuterMaxiter |
maximum number of allowed iteratons for outer loop, default to be 500 |
InnerMaxiter |
maximum number of allowed iteratons for inner loop, default to be 500 |
calpha |
Lagrangian parameter, default to be 10 |
b Estimated Intercept
alpha coeffiencents of the dual of MKL
weight Estimated between kernel weight
rho Estimated within kernel weight
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 | data(benchmark.data)
data.mkl=benchmark.data[[1]]
kernels=rep('radial',2)
sigma=c(2,1/20)
train.samples=sample(1:nrow(data.mkl),floor(0.7*nrow(data.mkl)),replace=FALSE)
degree=sapply(1:length(kernels), function(a) ifelse(kernels[a]=='p',2,0))
#Kernels.gen splts the data into a training and test set, and generates the desired kernel matrices.
#Here we generate two gaussisan kernel matrices with sigma hyperparameter 2 and 0.05
K=kernels.gen(data=data.mkl[,1:2],train.samples=train.samples,kernels=kernels,sigma=sigma,
degree=degree,scale=rep(0,length(kernels)))
C=0.05 #Cost parameter for DALMKL
K.train=K$K.train
K.test=K$K.test
# parameters set up
ytr=data.mkl[train.samples,3]
#Converts list of kernel matrices in to an array with is appropriate for C++ code
k.train=simplify2array(K.train)
k.test=simplify2array(K.test)
#Implement DALMKL with the hinge loss function
spicy_svmb1n=SpicyMKL(K=k.train,y=ytr, loss='hinge',C=C)
#Implement DALMKL with the hinge loss function
spicy_logistic=SpicyMKL(K=k.train,y=ytr, loss='logistic',C=C)#'
|
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.