- remove extra input parameter
f.criterion
fromtpr.dag.cv
,tpr.dag.holdout
,find.best.f
andcompute.fmax
: type of F-measure used to select the best F-measure is always the harmonic mean between the average precision and recall (f.criterion="F"
) and never the F-measure computed as average across examples (f.criterion="avF"
); - fix a minor bug in
tpr.dag.holdout
; - add warning checks in
tpr.dag.cv
andtpr.dag.holdout
; - improve some test cases and manual;
- add
build.scores.matrix.from.list
; - add
build.scores.matrix.from.tupla
; - add several test cases;
- streamline and lighten HEMDAG's hierarchical functions (namespace clearer and lighter);
- rename the following functions:
- htd-dag:
Do.HTD
-->htd.vanilla
;Do.HTD.holdout
-->htd.holdout
;
- obozinski heuristic methods:
heuristic.max
-->obozinski.max
;heuristic.and
-->obozinski.and
;heuristic.or
-->obozinski.or
;Do.heuristic.methods
-->obozinski.methods
;Do.heuristic.methods.holdout
-->obozinski.holdout
;
- gpav:
GPAV
-->gpav
;GPAV.over.examples
-->gpav.over.examples
;GPAV.parallel
-->gpav.parallel
;Do.GPAV
-->gpav.vanilla
;Do.GPAV.holdout
-->gpav.holdout
;
- tpr-dag:
TPR.DAG
-->tpr.dag
;Do.TPR.DAG
-->tpr.dag.cv
;Do.TPR.DAG.holdout
-->tpr.dag.holdout
;
- utility functions:
get.parents
-->build.parents
;get.parents.top.down
-->build.parents.top.down
;get.parents.bottom.up
-->build.parents.bottom.up
;get.parents.topological.sorting
-->build.parents.topological.sorting
;get.children.top.down
-->build.children.top.down
;get.children.bottom.up
-->build.children.bottom.up
;check.DAG.integrity
-->check.dag.integrity
;do.subgraph
-->build.subgraph
;do.submatrix
-->build.submatrix
;do.stratified.cv.data.single.class
-->stratified.cv.data.single.class
;do.stratified.cv.data.over.classes
-->stratified.cv.data.over.classes
;do.unstratified.cv.data
-->unstratified.cv.data
;do.edges.from.HPO.obo
-->build.edges.from.hpo.obo
;
- performance metrics:
AUPRC.single.class
-->auprc.single.class
;AUPRC.single.over.classes
-->auprc.single.over.classes
;AUROC.single.class
-->auroc.single.class
;AUROC.single.over.classes
-->auroc.single.over.classes
;compute.Fmeasure.multilabel
-->compute.fmax
;
- htd-dag:
- remove the following functions (no more needed):
Do.flat.scores.normalization
;Do.full.annotation.matrix
;
- improve manual;
- make HEMDAG's documentation clearer and less redundant;
- fix
stringsAsFactors
issue -- link;
- fix NAMESPACE notes in CRAN checks;
- add link to the GitHub repository
obogaf::parser
; - adjust link to read the docs;
- add
build.consistent.graph
;
- add some warning checks in functions that compute performance metrics;
- improve some graph utility functions;
- improve manual;
- improve tutorial on read the docs -- link;
- make namespace clearer;
- fix minor bugs;
- remove defunct functions;
- fix a minor bug in
Do.GPAV.holdout
; - improve package description;
- fix degenerate case in
precision.at.all.recall.levels.single.class
(labels are all negatives/positives); - fix degenerate case in
precision.at.given.recall.levels.over.classes
(labels in a fold are all negatives/positives); - fix degenerate case in
do.stratified.cv.data.single.class
(sampling of the labels with just one positive/negative); - add input variable
compute.performance
to the following high level functions:Do.TPR.DAG
andDo.TPR.DAG.holdout
;Do.HTD
andDo.HTD.holdout
;Do.GPAV
andDo.GPAV.holdout
;Do.heuristic.methods
andDo.heuristic.methods.holdout
;
- improve manual;
- add
lexicographical.topological.sort
;
- fix minor bugs;
- improve manual;
- precision-recall performance computed through
precrec
package:- add
precision.at.all.recall.levels.single.class
; PXR.at.multiple.recall.levels.over.classes
-->precision.at.given.recall.levels.over.classes
;
- add
- improve IO functions: the extension of the input or output file can be or plain text (
.txt
) or compressed (.gz
);
- fix minor bugs;
- improve manual;
- fix
CRAN
Package Check Results: remove unneeded header and define fromGPAV C++
source code
- add
GPAV
algorithm (Burdakov et al., Journal of Computational Mathematics, 2006 -- link); - Embed
GPAV
algorithm in the top-down step of the functionsTPR.DAG
,Do.TPR.DAG
andDo.TPR.DAG.holdout
; - Some functions have been defunct. To know the defunct functions just typing in the R environment:
help("HEMDAG-defunct")
;
- improve manual;
- add Alessandro Petrini as author for his contribution in writing the
C++
code ofGPAV
algorithm;
- various fixes from 2.1.2
-
improve performance metrics:
- add
compute.Fmeasure.multilabel
; - add
PXR.at.multiple.recall.levels.over.classes
; - all the performance metrics (
AUPRC
,AUROC
,FMM
,PXR
) can be computed either one-shot or averaged across folds;
- add
-
improve the high-level hierarchical ensemble functions:
- embed the new performance metric functions;
- add the parameter
metric
: maximization byFMAX
orPRC
(see manual for further details); - add some checkers (warning/stop messages) to make the library more user-friendly;
- improve manual;
- fix bug in
do.stratified.cv.data.single.class
;
-
add
TPR-DAG
: function gathering several hierarchical ensemble variants; -
add
Do.TPR.DAG
: high-level function to runTPR-DAG
cross-validated experiments; -
add
Do.TPR.DAG.holdout
: high-level functions to runTPR-DAG
holdout experiments; -
The following
TPR-DAG
andDESCENS
high-level functions were remove:Do.tpr.threshold.free
;Do.tpr.threshold.cv
;Do.tpr.weighted.threshold.free.cv
;Do.tpr.weighted.threshold.cv
;Do.descens.threshold.free
;Do.descens.threshold.cv
;Do.descens.weighted.threshold.free.cv
;Do.descens.tau.cv
;Do.descens.weighted.threshold.cv
;Do.tpr.threshold.free.holdout
;Do.tpr.threshold.holdout
;Do.tpr.weighted.threshold.free.holdout
;Do.tpr.weighted.threshold.holdout
;Do.descens.threshold.free.holdout
;Do.descens.threshold.holdout
;Do.descens.weighted.threshold.free.holdout
;Do.descens.tau.holdout
;Do.descens.weighted.threshold.holdout
;
NOTE: all the removed functions can be run opportunely by setting the input parameters of the new high-level function
Do.TPR.DAG
(for cross-validated experiments) andDo.TPR.DAG.holdout
(for hold-out experiments);
- improve manual;
- add
DESCENS
algorithm; - add Heuristic Methods
Max
,And
,Or
(Obozinski et al., Genome Biology, 2008 -- link); - add
tupla.matrix
function;
- improve manual;
- add link to the GitHub repository
HPOparser
(note: from version2.6.0
HPOparser
was changed inobogaf::parser
); - add
CITATION
file;