Tree fits are now also accepted from from the
partykit package (constparty objects), instead
of only from the rpart package.
Functions fusedTree and PenOpt now take
a new optional input argument symFusion. Setting this
argument to FALSE allows for asymmetric fusion, where omics
effects in nodes having similar predictions of the response (using only
clinical variables) will be shrunken less toward each other compared to
omics effects in nodes having more distinct predictions.
For memory efficiency, Matrix algebra is now conduced using
sparse matrices from the Matrix package when X
is high-dimensional.
Predict function does not require the test response by default anymore. Rather, it is an optional parameter.