Releases: skent259/mildsvm
Releases · skent259/mildsvm
v0.4.0
First version available on CRAN
Add ordinal methods to the package
- Add
omisvm()
for ordinal multiple instance support vector machine - Add
mior()
for multiple instance ordinal regression - Add
misvm_orova()
for MI-SVM reducing ordinal to binary one-vs-all classification - Add
svor_exc()
for support vector ordinal regression with explicit constraints
Other changes
- Breaking: change
generate_mild_df()
to a new interface - Breaking: change
mildsvm()
tomismm()
- Breaking: fix S3 method issue, affects
mi_df
andmild_df
methods parameter - Add
mi_df()
class and methods, includingas_mi_df()
- Add method for
mi_df
objects formisvm()
,cv_misvm()
and all new ordinal methods - Add
ordmvnorm
data for examples - Add print methods for
kfm_exact
,kfm_nystrom
,mild_df
,mior
,misvm
,mismm
,misvm_orova
,omisvm
,smm
,svor_exc
- Package now depends on R > 3.5.0, new imports of pillar, utils
- fix warning when
misvm()
has matrix passed - fix
.reorder()
ambiguity - pass lintr checks
- re-work internals for easier testing
v0.3.1
v0.3.0
v0.2.0
Major changes to the package API to follow typical R package conventions.
- The main modeling functions (
misvm()
,mildsvm()
, andsmm()
) now have three methods:- Formula method (i.e.
misvm(mi(y, bags) ~ x1 + x2, data = df, ...)
) - Data-frame method (i.e.
misvm(x, y, bags, ...)
) - Method for the
mild_df
class (I.e.misvm(mil_data, ...)
). This method often performs non-trivial aggregation or transformation sincemisvm()
andsmm()
work naturally on MIL data and supervised data, respectively.
- Formula method (i.e.
- Prediction on main modeling functions always returns a tibble with a single column depending on the
type
argument - Kernel feature maps functions are now organized as
kfm_nystrom()
,kfm_exact()
with abuild_fm()
method. - Update
MilData
class tomild_df
class, and improve the class methods and constructors. - Many internal methods removed and restructured.