svm.fs {penalizedSVM} | R Documentation |
Fits SVM with variable selection (clone selection) using penalties SCAD and L1 norm.
## Default S3 method: svm.fs(x, y, fs.method = "1norm", cross.outer = 0, lambda1.set, lambda2.set = NULL, calc.class.weights = FALSE, seed = 240907, maxIter = NULL,...) run.scad(x,y, lambda1.set=NULL, class.weights) run.1norm(x,y,k=5,nu=0, lambda1.set=NULL, output=1, seed=seed)
x |
input matrix with genes in columns and samples in rows! |
y |
vector of class labels |
fs.method |
feature selection method. Availible 'scad' and '1norm' |
cross.outer |
fold of outer cross validation, default is 0, no cv. |
lambda1.set |
set of tuning parameters lambda1 |
lambda2.set |
set of tuning parameters lambda2, not yet in use |
calc.class.weights |
calculate class.weights for SVM, default: FALSE |
class.weights |
a named vector of weights for the different classes, used for asymetric class sizes. Not all factor levels have to be supplied (default weight: 1). All components have to be named. |
k |
k-fold cross validation, default: 5 |
nu |
nu: weighted parameter
|
output |
0 - no output, 1 - produce output, default is 0 |
seed |
seed |
maxIter |
maximal iteration, default: not used yet |
... |
additional argument(s) |
The goodness of the model is highly correlated with the choice of tuning parameter lambda. Therefore the model is trained with different lambdas and the best model with optimal tuning parameter is used in futher analysises.
The Feature Selection methods are using different techniques for finding optimal tunung parameters By SCAD SVM Generalized approximate cross validation (gacv) error is calculated for each pre-defined tuning parameter.
By L1-norm SVM the cross validation (default 5-fold) missclassification error is calculated for each lambda. After training and cross validation, the optimal lambda with minimal missclassification error is choosen, and a final model with optimal lambda is created for the whole data set.
classes |
vector of class labels as input 'y' |
sample.names |
sample names |
class.method |
feature selection method |
cross.outer |
outer cv |
seed |
seed |
model |
final model
|
Natalia Becker natalia.becker at dkfz.de
Zhang, H. H., Ahn, J., Lin, X. and Park, C. (2006). Gene selection using support vector machines with nonconvex penalty. Bioinformatics, 22, pp. 88-95.
Fung, G. and Mangasarian, O. L. (2004). A feature selection newton method for support vector machine classification. Computational Optimization and Applications Journal, 28(2), pp. 185-202.
predict.penSVM
, svm
(in package e1071)
my.seed<- 123 train<-sim.data(n = 200, ng = 100, nsg = 10, corr=FALSE, seed=my.seed ) print(str(train)) # train SCAD SVM #################### # define set values of tuning parameter lambda1 for SCAD lambda1.scad <- c (seq(0.01 ,0.05, .01), seq(0.1,0.5, 0.2), 1 ) # for presentation don't check all lambdas : time consuming! lambda1.scad<-lambda1.scad[2:3] # # train SCAD SVM fit.scad<- svm.fs(x=t(train$x),y=train$y, fs.method="scad", cross.outer= 0, lambda1.set=lambda1.scad, seed=my.seed) # train 1NORM SVM ################ # define set values of tuning parameter lambda1 for 1norm epsi.set<-vector(); for (num in (1:9)) epsi.set<-sort(c(epsi.set, c(num*10^seq(-5, -1, 1 ))) ) # for presentation don't check all lambdas : time consuming! lambda1.1norm <- epsi.set[c(3,5)] # 2 params # train 1norm SVM # time consuming: for presentation only for the first 100 samples ## DON'T RUN : fit.1norm<- svm.fs(x=t(train$x),y=train$y, fs.method="1norm", cross.outer= 0, lambda1.set=lambda1.1norm, seed=my.seed) fit.1norm<- svm.fs(x=t(train$x)[1:100,],y=train$y[1:100], fs.method="1norm", cross.outer= 0, lambda1.set=lambda1.1norm, seed=my.seed)