SurvMKL: SurvMKL

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

Description

This function conducts SurvMKL for precomputed gramm matrices

Usage

SurvMKL(K, y, del, rho0, C = 0.5, lambda = 0.5, maxiter = 500, cri = 0.01)

Arguments

K	The multiple kernel cube (3-d array)
y	Vector of survival times
del	Indicator vector of whether an event occured (0 = no event, 1 = event occured)
rho0	Argmin Dual objective function
C	cost parameter is the loss function
lambda	tuning parameter for the elastic net. Lambda closer 1 corresponds to L1 and closer to 0 corresponds to L2 penalties.
maxiter	maximum number of allowed iteratons for outer loop, default to be 500
cri	change between to iterations is smaller than this, algorithms is considered to have converged, default to be .001

Value

alpha coeffiencents of the dual of SurvMKL

Examples

## Not run: 
library(RMKL)
library(kernlab)
data(Surv_data)
#Getting survival times in ascending order
ordtr <- order(Surv_data$time)
Surv_data_ordered = Surv_data[ordtr,]

xx = Surv_data_ordered[,1:2]
del = Surv_data_ordered$status
yy = Surv_data_ordered$time
if (!del[1]) {
  first1 <- which(del)[1]
  xx <- xx[-(1:(first1 - 1)), ]
  yy <- yy[-(1:(first1 - 1))]
  del <- del[-(1:(first1 - 1))]
  nn <- dim(Surv_data)[1] - first1 + 1
} else {
  nn <- dim(Surv_data)[1]
}

rho0 <- .001*(Surv_data$status - seq(0, 10, length.out = dim(Surv_data)[1]))
klist <- list(kernelMatrix(rbfdot(1), as.matrix(xx)),
              kernelMatrix(vanilladot(), as.matrix(xx)))
ktlist <- list(kernelMatrix(rbfdot(1), as.matrix(xx), as.matrix(Surv_data[,1:2])),
               kernelMatrix(vanilladot(), as.matrix(xx), as.matrix(Surv_data[,1:2])))
kk <- simplify2array(klist)
kkk <- simplify2array(ktlist)
modmkl <- SurvMKL(y = Surv_data$time, del = Surv_data$status, K = kk, rho =  
rho0, C = 0.005, lambda = 0.5, maxiter = 500, cri = .01)

## End(Not run)

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