PipeOpFDACor, PipeOpFDADerive,
PipeOpFDAExtract, PipeOpFDAIntegrate,
PipeOpFDAInterpol, PipeOpFDAScaleRange,
PipeOpFDASmooth, PipeOpFDAWavelets, and
PipeOpFDAZoom.PipeOpFDABsignal now uses the argument values
instead of index positions, which changes the features for grids other
than 1, ..., p.PipeOpFDAInterpol now determines the
"union", "intersect", and
"minmax" grids during training and reuses them during
prediction, instead of computing them from the prediction data.PipeOpFDARegister no longer fails during
prediction when the "srvf" method gets
fdasrvf::time_warping() arguments such as
penalty_method or centroid_type in
args.PipeOpFDAScaleRange now keeps the evaluator and
maps the domain to [lower, upper].PipeOpFDAScaleRange now stores its state under
$state$trafos.PipeOpFDAWavelets no longer supports irregular
functional data, where coefficients of curves with different lengths
were misaligned. Use PipeOpFDAInterpol first.PipeOpFPCA now uses the n_components
from training during prediction.PipeOps no longer forward
affect_columns to their backend functions.PipeOpFDAInterpol now reports which
grid, left, and right combination
is invalid instead of failing with an opaque assertion.PipeOpFDAInterpol now errors when
left and right fall outside the domain of a
functional column.tf version 0.5.0.PipeOpFDAWavelets no longer errors when
filter is a wt.filter object or a numeric
vector. Generated columns are now named
<column>_wav_<i> regardless of the filter.PipeOpFDAZoom now accepts a one-sided
begin or end instead of requiring both.as.data.table(mlr_pipeops) no longer errors with
object 'value' not found when mlr3fda is
loaded.PipeOpFDACatch22 for extracting the catch22
time series features from functional columns via
Rcatch22::catch22_all().PipeOpFDAIntegrate for extracting the
definite integral of functional columns via
tf::tf_integrate().mlr3fda to
mlr_reflections$loaded_packages to fix errors when using
mlr3fda in parallel.PipeOpFDADepth for computing the data depth
of functional columns via tf::tf_depth().PipeOpFDADerive for computing derivatives of
functional columns via tf::tf_derive().PipeOpFDAExtract gained the "sd"
feature for extracting the standard deviation.PipeOpFDAFourier is now several times
faster.PipeOpFDARegister for aligning functional
columns to a learned template via tf::tf_register().fuel task now correctly names the scalar
column h2o instead of h20.Mlr3Error and Mlr3Warning classes
for errors and warnings.PipeOpFDAFourier for extracting fast Fourier
transform features from functional columns.PipeOpFDABsignalPipeOpFDARandomEffectPipeOpFDATsfeaturesPipeOpFDAWaveletsPipeOpFDAZoomPipeOpFDACorPipeOpFDAScaleRangeInitial CRAN release.