Description Usage Arguments Value References See Also Examples
Does a k-fold cross-validation for kmr
, and returns a fitted KMR model, CV performance scores and optimal values for the regularization parameter lambda
.
1 2 3 4 5 |
x |
|
y |
|
kx_type |
Kernel type for observations as in |
kx_option |
Optional list of parameters for the observation kernel as in |
kt_type |
Kernel type for tasks as in |
kt_option |
Optional list of parameters for the task kernel as in |
lambda |
Sequence of (more than one) values for lambda that must be tested. Default is 10^(-5:5). |
type.measure |
Measure type for evaluating performance. Possible options are
" |
nfolds |
Number of folds for cross-validation. Default is 5. |
nrepeats |
Number of times the k-fold cross-validation is performed Default is 1. |
seed |
A seed number for the random number generator (useful to have the same CV splits). |
mc.cores |
Number of parallelable CPU cores to use. |
An object (list) of class "cv.kmr"
, which can then be used to make predictions for the different tasks on new observations, as a list containing the following slots:
... |
Outputs of a CV-fitted KMR model as in |
meanCV |
A matrix of CV performance scores of dim ntask x nlambda. |
bestlambda |
A vector of lambdas of length ntask, each corresp to the underlying min CV score. |
bestCV |
A vector of min CV performance scores of length ntask. |
lambda |
Lambda sequence against which a model is tested. |
type.measure |
Measure type for evaluating performance. |
Bernard, E., Jiao, Y., Scornet, E., Stoven, V., Walter, T., and Vert, J.-P. (2017). Kernel multitask regression for toxicogenetics. bioRxiv-171298.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 | # Data
ntr <- 80
ntst <- 20
nt <- 50
p <- 20
xtrain <- matrix(rnorm(ntr*p),ntr,p)
xtest <- matrix(rnorm(ntst*p),ntst,p)
ytrain <- matrix(rnorm(ntr*nt),ntr,nt)
ytest <- matrix(rnorm(ntst*nt),ntst,nt)
# Train with Gaussian RBF kernel for x and multitask kernel for t
cvmo <- cv.kmr(x=xtrain, y=ytrain, kx_type="gaussian", kx_option=list(sigma=100),
kt_type="multitask", kt_option=list(alpha=0.5), type.measure="mse")
# Plot
plot(cvmo)
# Predict
cvpred <- predict(cvmo, xtest)
# Evaluate
evalpred(cvpred, ytest, "mse")
|
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.