| 2fd339b | pfistfl | 11 April 2018, 15:19:41 UTC | Add basic tests | 11 April 2018, 15:19:41 UTC |
| 6bbc823 | pfistfl | 11 April 2018, 15:10:58 UTC | Add tsfeatures extractor | 11 April 2018, 15:10:58 UTC |
| 06cd031 | pfistfl | 11 April 2018, 14:28:09 UTC | Merge branch 'fda_lte' of github.com:mlr-org/mlr into fda_lte | 11 April 2018, 14:28:09 UTC |
| 0816dac | pfistfl | 11 April 2018, 14:27:50 UTC | Fix multiclass for FDA and fix Fourier features | 11 April 2018, 14:27:50 UTC |
| 3ba8db1 | smilesun | 11 April 2018, 09:04:52 UTC | add dtw to descrption | 11 April 2018, 09:04:52 UTC |
| 93b85fe | smilesun | 11 April 2018, 09:03:17 UTC | add hyper para set for DTW feature extraction | 11 April 2018, 09:03:17 UTC |
| db1df2c | pfistfl | 06 April 2018, 14:29:30 UTC | Fix defaults again, this is kind of ugly | 06 April 2018, 14:29:30 UTC |
| e57b50d | pfistfl | 06 April 2018, 14:03:41 UTC | Merge branch 'fda_lte' of github.com:mlr-org/mlr into fda_lte | 06 April 2018, 14:03:41 UTC |
| 9cefdc2 | pfistfl | 06 April 2018, 14:02:52 UTC | Fix metric for knn/np and kernel | 06 April 2018, 14:02:52 UTC |
| b467357 | Florian | 06 April 2018, 08:23:51 UTC | Fda lte fix parsets (#2243) * Allow paramsets for extractFDAFeaturesWrapper * Enable tuning of extractFDAFeaturesMethods * Extend unittests * Move Bsignal and add paramset * Rename fdafgam to fgam | 06 April 2018, 08:23:51 UTC |
| 14d379f | Florian | 04 April 2018, 15:19:47 UTC | Fda lte fix parsets (#2239) * Allow paramsets for extractFDAFeaturesWrapper * Enable tuning of extractFDAFeaturesMethods * Extend unittests | 04 April 2018, 15:19:47 UTC |
| 40407a5 | smilesun | 04 April 2018, 09:41:32 UTC | rename learners | 04 April 2018, 09:41:32 UTC |
| 385f664 | smilesun | 04 April 2018, 09:35:45 UTC | fgam refine | 04 April 2018, 09:35:45 UTC |
| ffcee0f | smilesun | 05 March 2018, 17:34:57 UTC | make a better version of fgam regr | 05 March 2018, 17:34:57 UTC |
| e6f53c3 | Florian | 21 February 2018, 16:57:24 UTC | Fda pull2 fix extract fda features (#2176) * Init fda with matrix * More unit tests, add some convenience * Colname prints are ugly for already provided matricies * Add functionals to task.desc.feat * getTaskData changed * Add unittests for functionals = TRUE flag in changeData * Add examples for makeFunctionalData * Fix target. extra, now returing message when converting * More unittests, fix tiny bug * Fix unittests for orderedfactors, remake tasks to update task.desc * datasets definition changed * Add learners and tests * test objects and tasks for fda structure added * Change desc, unittests etc. * fdaknn implemented, unittests pass * SUpport probabilities for fdaknn * Silence lintr * Fix typo in unittests * Silence message * Fix resampling and add probs / unittests for prrobs for fdaknn * ROC and tunethreshold tests * Fix subsetTask for functionals * generateCalibration test * Improve FDBoost, Needs fixed by smilesun * added tests for Threshold calibration and ROC * Silence lintr 1/2 * Fix lintr and add more unittests * Silence lintr 2/2 * fdaregr and fdaclassif benchmarks for fda and non-fda lrns * Fix newlien * Fix and add unittests, add more informative error messages * Extend documentation, add unittest * Update docu for makeTaskData * Added learners, already fixed np, glm and knn * Rename learner files * change kernel learner * Update fdakernel, still something missing * Add probs, fix unittests * Fix suggestions by Bernd 1/2 * Fixed learners * Fix unittests for new getTaskData * Remake fda datasets * Remake gunpoint.task * Fix TaskDesc and remake datasets * hand task instead of regr|classif in order to allow for fda learners * Fix problems introduced by vcapply and ordered factors * Fix unittests for new makeFunctionalData * Rename fgam, move shapelets to todo-files * Rename shapelet learner files, start fixing regression learners * Fix documentation * Restart travis * Add FeatureExtraction * Fix fda feature extractors * Fix some unittests for featureExtraction * Remove reference to templates * restart travis * Fixed unittests and some residual bugs from porting extractFDAFeatures * Fix extractFDAFeatures * Add note to shapelet learner * Add extractFDAFeatures * extractFDAFpca * Start fixing multires features * Add extractMultiResFeatures * Fix multiresolution features and some unittests * fix FDboost (#1901) * Fixed unittests and FDboost * Speedup fdBoost, start fixing fgam * Move fgam learner to todo-files * Update News * Add phoneme multiclass fda task * Add single.funcitional property * Update news, lapply instead of sapply * Restart travis * Fix merge conflicts 2/2 * Fix FDboost learner, can now deal with ONLY numerics * Fix indents to restart travis * Fix lintr * Silence lintr 2/2 * Update documentation to pass R CMD check * Fix docu 2/2 * Fix lintr again * Fix typo * Remove man again * Revert namespace * ... * ... * ... * Rename classif learners to fdausc.name * ... * Rename regr learners for functional data * Fix missing update in kernel * ... * Update unittests for new learner names and dfcols * Update FIXMEs in tests * ... * ... * ... * ... * ... * ... * makeFunctionalData now allows NULL and empty list * ... * extractFunctions into Methods * Rename task in example and reExtract to reextract * Fix docs * Fix and restructure tests * Fix unittests for extractFDAFeatures * Delete empty test files * Fix unittests and reextract calls * Update makeFunctionalData * Docu for getFunctionalFeatures, fixed bugs, more asserts * Unittests for getFunctionalFeatures * Extend unittests! * Update docs * Add docu for task, update featureless learns, unittests for featureless lrns * Fix tests * feat.methods can now be multiply specified * dont export getFunctionalFeatures * NEWS * ... * ... * ... * ... * ... * ... * Fix # 1928 * Fix #1930 * Fix #1933, cols -> col, revert new param for now * Fix cols -> col * Fix typo, add comment * assert Namedlist * Fix missing family, add comment * Add asserts and enhance docu for multires feature extraction * Better asserts * Fix docs and example for extractFDAFeatures * Fixed comments by Bernd * assertCharacter -> assertString * Make description better * Default as last * Re-add example * First fix for wavelets methods, remove unccessessary FIXME * Fix missing objects, adjust unittests * Update documentation * Update docs * Fix newlines / merge artifacts * Re-add and fix example * Re-enable unittests, changed feat extract output colnames * Fix unittests and undocument missing param * Delete duplicated function * space added | 21 February 2018, 16:57:24 UTC |
| 91f0f9f | smilesun | 21 February 2018, 16:45:29 UTC | change code to backdot to solve conflit | 21 February 2018, 16:45:29 UTC |
| 8d33569 | smilesun | 21 February 2018, 16:35:15 UTC | Merge branch 'master' into fda_lte | 21 February 2018, 16:35:15 UTC |
| f20daef | Patrick Schratz | 19 February 2018, 08:40:17 UTC | update man-roxygen (#2174) | 19 February 2018, 08:40:17 UTC |
| 91d3128 | Michel Lang | 16 February 2018, 20:43:31 UTC | simplified interactive testing | 16 February 2018, 20:43:31 UTC |
| 009b2cd | Patrick Schratz | 16 February 2018, 16:30:56 UTC | fix documentation issues in man-roxygen (#2172) | 16 February 2018, 16:30:56 UTC |
| c45b6cc | Steve Bronder | 16 February 2018, 16:28:04 UTC | Add myself as ctb for windowing CV (#2171) * Add myself as ctb for windowing CV * ... * Update DESCRIPTION | 16 February 2018, 16:28:04 UTC |
| 386939e | bommert | 15 February 2018, 09:57:45 UTC | add ranger filters (#2165) | 15 February 2018, 09:57:45 UTC |
| 6069b98 | Travis | 13 February 2018, 22:05:00 UTC | update auto-generated documentation [ci skip] | 13 February 2018, 22:05:00 UTC |
| 26d9f23 | Lars Kotthoff | 13 February 2018, 20:56:13 UTC | NEWS for #2164 | 13 February 2018, 20:56:13 UTC |
| 1652e38 | Michel Lang | 13 February 2018, 20:55:23 UTC | Regenerate example data (#2164) * add spam.task, rebuilt data * rebuilt data using devtools * nolint | 13 February 2018, 20:55:23 UTC |
| 28b6ff4 | Travis | 13 February 2018, 11:53:21 UTC | update auto-generated documentation [ci skip] | 13 February 2018, 11:53:21 UTC |
| 72297a7 | Alexander Engelhardt | 13 February 2018, 10:49:23 UTC | [WIP] Rewrite measure tests for new mldr package functions (#2161) | 13 February 2018, 10:49:23 UTC |
| 6e6268d | Lars Kotthoff | 12 February 2018, 17:12:49 UTC | NEWS for #2163 | 12 February 2018, 17:12:49 UTC |
| c52b9cc | Alexander Engelhardt | 12 February 2018, 17:11:07 UTC | Rename measureMultiLabelF1 to measureMultilabelF1 (lower case L) (#2163) * Rename measureMultiLabelF1 to measureMultilabelF1 (lower case L) * keep deprecated notation | 12 February 2018, 17:11:07 UTC |
| ef52d41 | Patrick Schratz | 09 February 2018, 16:25:40 UTC | Enable roxygen2 markdown support globally (#2143) * enable roxygen markdown support * disable markdown support for conflicting functions * convert to markdown syntax * fix several problems after markdown converison (tbc) * manual fixes of Rd problems * replace [`character(1)`] with (`character(1)`) * replace [`character`] with ([character]) * replace [`numeric`] with ([numeric]) * multiple replacements * intermediate commit * various changes inlcuding man-roxygen/ * last changes?! * change roxygen note manually * remove all () | 09 February 2018, 16:25:40 UTC |
| c08bb00 | Zachary M. Jones | 08 February 2018, 12:58:31 UTC | partial dependence plot fix and test for binary classification (#2153) found the issue while updating the tutorial, added two tests that would have caught the tutorial bug and another one that could have occured | 08 February 2018, 12:58:31 UTC |
| a15d5c7 | Julia Fried | 08 February 2018, 11:11:11 UTC | linked cheatsheet (#2162) | 08 February 2018, 11:11:11 UTC |
| 9865cea | Lars Kotthoff | 01 February 2018, 17:18:26 UTC | NEWS for #2121 | 01 February 2018, 17:18:26 UTC |
| f3ee600 | Steve Bronder | 01 February 2018, 17:15:53 UTC | merge growing window and fixed window cross validation (#2121) * Create functions for growing and fixed window cross validation Update docs for resample to include horizon, initial.window, and skip Fix doc for skip * update to commens from Lars * Names changed to GrowingWindowCV and FixedWindowCV, add full stops to all messages, and gives a warning when the last value is not included * Now window resamples can accept values < 1 or values > 1. Better warning for missing end values and included tests to check domain of values and warning is thrown * Fixes names for skip and horizon to be consistent * Fix print so that it accepts fractions * Make windowing function in helpers for growing and fixed window CV | 01 February 2018, 17:15:53 UTC |
| a0f5ad7 | Zachary M. Jones | 30 January 2018, 23:05:22 UTC | bugfix and new tests for generatePartialDependenceData (#2148) jacobian estimation bug which was hidden in tests because the multiclass tests where the jacobian was estimated use the iris data where the number of classes matched the n[1] arg tests are added that would have caught it and n[1] was changed to equal 4 as well as the actual bug being fixed | 30 January 2018, 23:05:22 UTC |
| 6e0768a | Travis | 30 January 2018, 18:55:26 UTC | update auto-generated documentation [ci skip] | 30 January 2018, 18:55:26 UTC |
| 3321524 | Zachary M. Jones | 30 January 2018, 17:58:43 UTC | move mmpf to suggests (#2145) | 30 January 2018, 17:58:43 UTC |
| 06fb95e | Travis | 26 January 2018, 19:34:25 UTC | update auto-generated documentation [ci skip] | 26 January 2018, 19:34:25 UTC |
| fa11d6d | Lars Kotthoff | 26 January 2018, 18:32:16 UTC | NEWS for #2134 | 26 January 2018, 18:32:16 UTC |
| 2ebd2f2 | Patrick Schratz | 26 January 2018, 18:31:27 UTC | New spatial implementation approach (#2134) * render html * remove showHyperPars --> getParamSet * revised imbalanced classification * proofreading, mention predict.threshold option * updated html imbalanced classification problems * update 'Visualizing the prediction' for classification (fixes #241) * updated html cost-sensitive classification * revised ROC analysis, threshold tuning moved to its own section * render html * cleanup, links, typos * make learner table more compact * render tutorial * tutorial update for roc_analysis * update tutorial with generateROCRCurves * update tutorial with generateLearningCurve * integrated learners table alignment (Issue #268) * replace getFilterValues by generateFilterValuesData * reworked section about benchmark experiments; examples for nested resampling were removed in favor of a separate section about nested resampling * use the new structure for pages config in the mkdocs.yml * generateThreshVsPerfData * tutorial: parallel: simply add 'mlr.' as prefix to level names to be explicit, fixes #178 * tutorial: parallel: explain exporting for custom learners * generate html with mlr version 2.4 * update tutorial with plot/generate and ggvis * typos and macro fixes * proofreading and cleanup; mention function getTaskSize and option fix.factors.prediction of makeLearner * generate html * remove old tutorial stuff and add 404 to direct to new tutorial * save old structure (for google) and forward * symlinks wont work * forward to devel version directly * add example site * docs/news * update docs * update docs * update docs * update travis * docs * update travis * build cache * build cache * dont run tests * add tic.R * only run script in deploy stage * next deploy attempt * only deploy * install lintr from github * force commit and do not build locally * add new tutorial chapters * add more tutorials * update gitignore * typos * add ggplot2 * update docs * update docs * add images, update internal links * require pander * install forked pkgdown * fake change * remove docs from gitignore * Deploy from Travis build 42 [ci skip] Build URL: https://travis-ci.org/pat-s/mlr/builds/311728468 Commit: eea8e00ed04d19879bd040ac01d733c382b556d3 * typo * delete docs manually to trigger rebuild * Deploy from Travis build 44 [ci skip] Build URL: https://travis-ci.org/pat-s/mlr/builds/311981995 Commit: dd4129cc0cb8ea757051035fae2e419cebe3f79b * test cross-references for internal functions * Deploy from Travis build 46 [ci skip] Build URL: https://travis-ci.org/pat-s/mlr/builds/313901337 Commit: 16ee87a19ffff465b76ac3b59babc3c544a5f502 * fix travis error * try deploying to gh-pages * build again * add some orcid ids * update visualization.rmd * fix DESCRIPTION * update articles * fix link for resample * delete docs * update articles * add mlr logo * add logo as favicon * enable automatic pkgdown linking * - switch theme to cosmo - enable slack icon * add blog, stackoverflow and slack icon * add ORCID * update articles * require imagemagick * update articles * fix link for logo * update custom_learners.md * update custom_measures.Rmd * update articles * update articles * update impute.Rmd * update articles * update multilabel.Rmd * update nested_resampling.Rmd * fix * update articles * update articles * update predict.Rmd * update preproc.Rmd * update resampling.Rmd * update roc_analysis.Rmd * update Task.Rmd * update tune.Rmd * update articles * update png paths, add pdf section * install here * add git deploy, fix here installation * install here explicitly * update tutorials * README: add mlr-tutorial instructions * rename to _pdf_wrapper.Rmd so it gets ignored by pkgdown * build pdf version * install bookdown * update pdf paths * here also for pdf version * build mlr * add previous mlr-tutorial versions * update README * fix paths * explicitly install mlr from github (?) * install pander * update README, add pdf images, add cheatsheet * fix paths * set image width relative to output format * fix typo * add cran/devel switch * require qdap * update paths * change from 2.12 to devel/release * change from 2.12 to devel * build html and pdf in two different stages * build as matrix instead of stages * fix tic installation * update links * build cache... * build again * update refs * harmonize path of release version * move to old * update paths * fix paths * fix paths * try installing ghostscript to suppress error for pdf building * update release docs * try to get git committing * fix links in docs * reinstall rcppziggurat * clean up * fix links * readd gsl packages * only git push on success * update auto-generated tutorial pdf version [ci skip] * update urls * update auto-generated tutorial pdf version [ci skip] * rebase * try explicit toc with depth 4 * update tic.R * update auto-generated tutorial pdf version [ci skip] * change font to Fira Code * update auto-generated tutorial pdf version [ci skip] * fix tic.R * update auto-generated tutorial pdf version [ci skip] * change fonts * update auto-generated tutorial pdf version [ci skip] * body font: Fira Sans new font size and width for navbar * update auto-generated tutorial pdf version [ci skip] * correct placement of search bar * update auto-generated tutorial pdf version [ci skip] * - add <title> for each article - render output in blocks and not single lines * remove dev=svg * update auto-generated tutorial pdf version [ci skip] * fix #2125 * add test for nested SpCV * change from 'spatial' (lgl) to 'coordinates' (df) * update task data * spatial -> coordinates * remove unnecessary test * update initialization of task in test helpers * update bc.task.spatial data set * remove all the hackish stuff :-) * revert inital workaround needed for prior spatial approach * remove requirement of coordinates argument in ResampleInstances * update Task doc * add spcv tests * lintr * check for correct structure of coordinates argument * update Task doc * remove checks as they are already existent in makeTask * remove parallelization from tests * move spcv tests into own file * regenerate tasks * update tests with new data * update inst/makeData * update changeData arguments * update datasets.R * try solving permission change issue * delete pkgdown stuff * revert file permission changes * remaining file permission issues * remaining file permission issues * update datasets.R * update tests with new data * revert citation change * revert makeData.R change * check if task exists with coordinates * check with is.data.frame * fix if condition in ResampleInstances * fix rowNames of spatial.task * require the same row names for coordinates and data * update spcv checks * update rownames of spatial.task * fix if condition * move nrow coodinates check to Task.R * remove else statement and do subsetting of coordinates in subsetTask * clean citation * remove unnecessary test * lintr * indention * minor * Update Task doc and check if coordinates df has two columns * - require at least two dimensions - remove row name requirement * < not <= | 26 January 2018, 18:31:27 UTC |
| 54dea48 | Travis | 25 January 2018, 18:35:23 UTC | update auto-generated documentation [ci skip] | 25 January 2018, 18:35:23 UTC |
| c1d0727 | Lars Kotthoff | 25 January 2018, 17:42:23 UTC | NEWS for #2041 | 25 January 2018, 17:42:23 UTC |
| f09ee72 | Zachary M. Jones | 25 January 2018, 17:39:54 UTC | mmpf integration in generatePartialDependenceData (#2041) - speed improvements from using mmpf package and rewrite of generate function - remove centering - new functions for derivative estimationa - remove functional anova | 25 January 2018, 17:39:54 UTC |
| deb4a87 | Travis | 25 January 2018, 10:09:08 UTC | update auto-generated documentation [ci skip] | 25 January 2018, 10:09:08 UTC |
| f6a5b17 | ja-thomas | 25 January 2018, 09:03:38 UTC | NEWS for #1837 | 25 January 2018, 09:03:38 UTC |
| d2f52de | Bernd Bischl | 25 January 2018, 09:02:13 UTC | adds add mxnet (#1837) * ... * prefix mxnet functions * Add updated RLearner and tests (#1950) * Rename param validation.set to validation.ratio * Split dropout parameter into one parameter for every layer * Use makeResampleDesc for sampling to use stratification * Add comments concerning dimensionality of conv.data.shape * Use correct check for NULL * Comment call of mx.symbol.flatten for fully connected layers * Document predict function * Move testProbWithTol and testProbParsetsWithTol to helper_helpers * Adjust tests for altered dropout parameters * Fix lintr errors * Simplify parameter set of mxff and adjust tests * add mxnet to suggests * mxnet regression * add se estimation for mxnet * lint lint lint * Explicitely pass array.layout to mx.mlp in tests to avoid warning * magic number * Update RLearner_regr_mxff.R * Add batch normalization * Add conditionality to batch normalization parameters * Add global param for batch normalization * Update documentation of mxff * Set tunable to FALSE on hyperparameter conv.data.shape * Update prediction of RLearner_regr_mxff for predict.type se * Add basic unit test for RLearner_regr_mxff * lint | 25 January 2018, 09:02:13 UTC |
| 54ed2c8 | mb706 | 23 January 2018, 21:15:31 UTC | lintr parsed := correctly (#2137) | 23 January 2018, 21:15:31 UTC |
| a6465b1 | Lars Kotthoff | 18 January 2018, 21:15:17 UTC | fix ranger tests (#2135) | 18 January 2018, 21:15:17 UTC |
| 7541988 | mb706 | 21 December 2017, 17:58:50 UTC | Fix 2118 getRRPredictionList with predict.type se (#2119) * handling 'se' * adding test * lintr | 21 December 2017, 17:58:50 UTC |
| ebb5ba0 | Travis | 15 December 2017, 17:42:58 UTC | update auto-generated documentation [ci skip] | 15 December 2017, 17:42:58 UTC |
| 5028093 | Lars Kotthoff | 15 December 2017, 16:44:46 UTC | NEWS for #2106 | 15 December 2017, 16:44:46 UTC |
| 6894f14 | Alexander Engelhardt | 15 December 2017, 16:43:35 UTC | calculateConfusionMatrix() gets a new parameter set = "test" | "train" | "both" (#2106) * change braces to brackets (was typo) * Add argument 'set' to calculateConfusionMatrix() * lintr * lint II * add test for calculateConfusionMatrix set argument * lintr III * make calculateconfusionMatrix() raise an error when confusion matrix is empty * review changes | 15 December 2017, 16:43:35 UTC |
| eaf41b7 | Travis | 14 December 2017, 17:44:39 UTC | update auto-generated documentation [ci skip] | 14 December 2017, 17:44:39 UTC |
| e760e4f | Jakob Richter | 14 December 2017, 16:27:36 UTC | tuneMBO did return negative performance values when maximizing (#1955) * use smoof minimize property * more tests * remove forces, add tests * documentation * fi * linting | 14 December 2017, 16:27:36 UTC |
| 6870986 | Lars Kotthoff | 14 December 2017, 16:22:08 UTC | NEWS for #2104 | 14 December 2017, 16:22:08 UTC |
| a2e945c | bommert | 14 December 2017, 16:20:37 UTC | Plot tune multi crit result (#2104) * add information about x space in plotTuneMultiCritResultGGVIS * fix seeding for new version of rfsrc (#2039) * add citations for papers in readme, see #1694 (#2040) * Fix issue #2047 and refix #838 (#2049) * Fix issue #2047 and refix #838 * linting * linting failed? * extend ranger hyperpars (#2058) * extend ranger hyperpars * default according to help is ignore * should be integer * Added @family filter to listFilterMethods (issue #1676) (#2045) * Added @family filter to listFilterMethods (issue 1676) * Revert "Added @family filter to listFilterMethods (issue 1676)" This reverts commit 7ae1ecd259f5f0e6eba77d5ceafb2b3110535aaf. * Add @family filter to listFilterMethods and makeFilter * Add faceting argument to plotROCCurves (#2009) * add faceting argument to roc curves * change default, tests and cleanup code * use correct && and || in if statements * replace sapply * logical result inside of apply * update auto-generated documentation [ci skip] * Update link in README.md [ci skip] * make information about points optional * code improvement and trafo * style improvement * code improvement | 14 December 2017, 16:20:37 UTC |
| 1c92306 | Michel Lang | 14 December 2017, 11:04:38 UTC | seeds for batchmark test | 14 December 2017, 11:04:38 UTC |
| c0daafb | Travis | 13 December 2017, 21:28:49 UTC | update auto-generated documentation [ci skip] | 13 December 2017, 21:28:49 UTC |
| ae08a46 | Bernd Bischl | 13 December 2017, 20:21:40 UTC | fix tunethreshold with MBO, fixes #2002 (#2013) * fix tunethreshold with MBO * ... * improve test * travis dump logs * remove forgotten parameter | 13 December 2017, 20:21:40 UTC |
| 0480872 | mb706 | 13 December 2017, 20:20:50 UTC | Fix costsens tasks (#2110) * Export getTaskCosts.CostSensTask * test for cost presence * regenerating costiris.task | 13 December 2017, 20:20:50 UTC |
| 1887c21 | Travis | 13 December 2017, 19:56:23 UTC | update auto-generated documentation [ci skip] | 13 December 2017, 19:56:23 UTC |
| 0c0bbe7 | mb706 | 13 December 2017, 18:17:46 UTC | CPO: Composable Preprocessing Operators (#1827) * Introducing Composable Preprocessing Objects. * ParamSet syntactic sugar * Make git ignore emacs temp files * Bugfixes in ParamSetSugar * Automatically generate function from braced expressions * lintr fixes * Creation of CPOObject * Nice printing * CPOObject concatenation * CPO composition * Composition operator also for attachment * wrapping seems to work now * some experiments * Some reorganizing * Implemented CPOFunctional, against all odds. Probably full of bugs still. * Reorg: Organize CPOObjectBased, CPOFunctional the same * Bugfixes * CPOObject now doesn't need to return 'control', just create it. * lintr * ParamSetSugar test * Indentation * Testing most of CPO, excluding ParamSet feasibility checks * Bugfixes, found through tests * setHyperPars: assert uniquely named parameters * Testing hyperparameter feasibility * Test parameter feasibility * Testing actual data transformation * Testing CPO trafo functions * Testing requirement handling * Forgive absence of parameters with unfulfilled requirements * Requirement handling when changing ID * Repair global var problems in S3 methods in cpo tests * lintr * Application operator * Corrected copy-paste caused typo * Inform user when he forgets to construct CPO * Documentation * Make R CMD check --no-test happy * lintr doesnt recognize CPO function definitions as functions * Retrafo set / access functions * paramSetSugar parameter pss.* now have dot prefix for R param matching reasons * retrafo() machinery * Functional CPO now uses retrafo() * Roxygenise * Tests work again * Bugfixes * More informative error messages * More informative error messages * Testing for error handling * Embarrassing! * static analyzer safe paramSetSugar * Using NA instead of dot to indicate missing parameter * Cleaning up documentation * Documentation fixes * lintr * Turn chain of preprocs into list, and assemble list into chain * roxygenize * Put common CPO test objects into helper_cpo.R * Refactor chainung and un-chaining * Chaining, unchaining of object based retrafos * use 'predict' to apply retrafos * lint * Adding get / set retrafo state functionality * Adding get state and makefromstate for object based * Cleaning up CPO object based * Testing for retrafo state * Cleaning up CPOFunctional * Adding get state and makefromstate for functional based * lintr * R CMD check * Small test correction Evidently I should clear my .GlobalEnv before running tests. * small comment change * retrafo assignment now checks for type, not function * Adding properties parameters * Added docu, todo * Starting task shape verification things * Get format check its own file * Auxiliary files reorg * changed object based callCPO[Re]Trafo, need to propagate the changes now * One more step towards properties & data shape checking * Tests pass again * cleaning up * Make lint approve of TODOs temporarily * Tests first half of target type functionality * checkLearnerBeforeTrain: Wrong error message when unordered not supported * lintr * NOOP * Finished datasplit tests * Most property tests are done * get CPO from learner * Ported properties and datasplit to functional * Tests pass * roxygenise * Make tests faster * travis timeout ++ * Travis timeout +++ * Rewritten CPO core. A beauty to behold! This removes 'makeCPOFunctional' and 'makeCPOObject' and replaces them both with 'makeCPO'. * Travis timeout ++++ * Added 'factor', 'ordered', 'onlyfactor', 'numeric' datasplit Numeric splits also support matrix instead of data frame * Introducing NULLCPO, the neutral element of the CPO monad * Starting targetbound CPO * Targetbound CPO Task conversion backend * static code analyser found bugs * Making big steps towards target CPOs * to-do list, travis timeout ++ again * Pretty much done with target-bound CPO * Introducing stateless CPOs * is.nullcpo * Completing stateless * Roxygenise * Tests pass * lintr * stateless trafo-less CPO * ShapeInfo printing * Nicer ShapeInfo printing * More generics for getting CPO information * Multiplexer, Applicator * Renaming test files * Split up test_cpo_datasplit into *_datasplit and *_properties * Checking par.vals availability at the right places * Proper datasplit numeric / factor / etc handling * New tests * Datasplit numeric, factor, ordered, onlyfactor finally seem to work * Finished datasplit tests * summary bug * example CPOs handle DFs containing non-numeric columns * repair summary * Accept NA vector length * Accept character vectors for discrete character params * cpoSelect CPO * cpoSelect Params Reorg * cpoCbind * Check more rigorously that CPOs don't get called too often. * Test cpoCbind with tasks * listCPO * Testing concrete CPOs so far * Fix retrafo column name test * fix.factors * dummy encoder * Column selection by name * invert option for cpoSelect * Starting to implement affect.* * Interpreting subset * Fixing some bugs, implementing some tests, for affect subset * Don't print meta-params for CPO constructors * Tests for affect.* done * Collect meta-CPOs in a separate file * Finishing cpoMeta and its tests * lintr * Export cpoMeta * Fix summary bug * Export some functions I forgot to export * Adding jupyter vignette * Adding html rendered version of vignette * update .gitignore * Compact html vignette * Fixing test bug for impute * Impute CPO + tests * Adding specialised CPO imputers * CPO Imputers get their own file * lintr * Test that dummys are not created when the flag says so. * Updating Vignette * checkMeasures: instead of missing(), use NULL * Feature Filters * Introducing applyCPO: apply a CPO to a Task / df * Introduce composeCPO: composing two CPOs * Introducing attachCPO: Attaching a CPO to a learner. * Adjust properties of imputers that can only handle certain types * Forgot export * filter features now only operate on the columns of the right type * Constant Feature Remover CPO * CPO for fixing factors * cpoDummyEncode works much better now * MissingIndicators CPO * Repairing checkMeasures * Better travis check * cpoCbind bugfixes * bugfix * Vignette updates * use cases ipynb * Recursive application of CPO fix * roxygenise * Avoid warnings when load_all-ing mlr * Now possible to specify packages associated with CPOs * Bugfix * Fix blackboost bug TODO: report this * Revert "Avoid warnings when load_all-ing mlr" This reverts commit ba2cf735ea17129a1a2e221c5da54d6a05a96f16. Reverting this because I made an extra pull-request * Revert "Fix blackboost bug" This reverts commit 9e1007fd526f4bc090fd62231b784e30ed30c7fe. Reverting this b/c I made an extra PR * Fix imputation of empty df bug * cpoScale fix for 1-column data * cpoApplyFun * cpoRangeScale * cpoProbEncode * Impact encoding * Adding new CPOs * rename cpoRangeScale -> cpoScaleRange * More natural handling of 'stateless' cpo * .retrafo.format added, with new 'combined' option * optionally only export subset of parameters * fix export * removing 'CPOS3Primitive' class, as ordered * removing 'CPOS3Constructor' class, as ordered * removing 'CPOS3RetrafoPrimitive', as ordered * simplifying class structure, as ordered * simplifying class structure further * done simplifying class structure * makeCPO documentation * Handle ID correctly for non-exported values * Pretty printing; Bugfixes * A few new CPOs * cpoSpatialSign * tuning CPO test * tuning CPO test II * bugfixes * roxygenise * Vignette: Examples, CPOs, Construction * cleaning up vignette * vignette html export * Bugfix * Tuning vignette * vignette * exporting necessary things * deleting superfluous files which are now in mlrCPO * forgot necessary function * Export makeBaseWrapper * Removing paramSetSugar * Roxygenise * Exporting changeData * Removing the last few bits from before the cpo-mlr-split * Export checkPredictLearnerOutput for mlrCPO * add 'keywords internal' to internal use only functions * roxygenise * overlooked while merging * %% | 13 December 2017, 18:17:46 UTC |
| d0c5d95 | mb706 | 13 December 2017, 17:46:07 UTC | set na.rm to TRUE for variance feature filter (#1938) * set na.rm to TRUE for variance feature filter * add unit test * lintr | 13 December 2017, 17:46:07 UTC |
| aec524c | Lars Kotthoff | 13 December 2017, 17:21:48 UTC | fix confusion matrix bug with different factor levels for response and truth (#2033) * fix confusion matrix bug with different factor levels for response and truth * fuck you lintr-bot | 13 December 2017, 17:21:48 UTC |
| 07ec916 | ja-thomas | 13 December 2017, 17:19:56 UTC | fix for constant predictions (#2061) | 13 December 2017, 17:19:56 UTC |
| 6c2dbb1 | Christian Scholbeck | 13 December 2017, 17:16:22 UTC | - changed se.method of regr.randomForest to 'sd' (#2095) - changed documentation of regr.randomForest accordingly | 13 December 2017, 17:16:22 UTC |
| 0e7a475 | Alexander Engelhardt | 13 December 2017, 16:55:02 UTC | add tests for calculateConfusionMatrix (#2089) | 13 December 2017, 16:55:02 UTC |
| fa29f45 | mb706 | 13 December 2017, 16:49:43 UTC | Fix task data (#2107) * Fixing surv task data files for problem #2010 * renewing tasks that differed from what changeData() gives * test for changed data format | 13 December 2017, 16:49:43 UTC |
| a49905d | Patrick Schratz | 12 December 2017, 04:28:29 UTC | add contributer | 12 December 2017, 04:28:29 UTC |
| 1c65e26 | Travis | 11 December 2017, 23:10:16 UTC | update auto-generated documentation [ci skip] | 11 December 2017, 23:10:16 UTC |
| 6c56e29 | Lars Kotthoff | 11 December 2017, 22:12:53 UTC | NEWS for #2100 | 11 December 2017, 22:12:53 UTC |
| 75e14de | mb706 | 11 December 2017, 22:08:55 UTC | Missing levels suggestion (#2100) * alternative missing levels removal * fixing bugs * fixing the only places that rely on the old .subset * task.desc saved for model differs if subset given * @mllg suggested changes | 11 December 2017, 22:08:55 UTC |
| b531936 | mb706 | 11 December 2017, 17:02:16 UTC | Fix 2092 no check data in cluster task (#2103) * Testing for construction with check.data = FALSE * Fixing #2092 | 11 December 2017, 17:02:16 UTC |
| 269176c | Travis | 11 December 2017, 10:19:21 UTC | update auto-generated documentation [ci skip] | 11 December 2017, 10:19:21 UTC |
| 0fb6136 | Jakob Richter | 11 December 2017, 09:24:15 UTC | Update NEWS.md: makeTuneMultiCritControlMBO [ci skip] | 11 December 2017, 09:24:15 UTC |
| 8865d5d | bommert | 11 December 2017, 09:20:14 UTC | Tune mbo multi crit (#2082) Enables mlr to use multicriteria tuning with mlrMBO | 11 December 2017, 09:20:14 UTC |
| 1ecfe27 | mb706 | 10 December 2017, 20:14:48 UTC | giving surv task the surv type (#2102) | 10 December 2017, 20:14:48 UTC |
| dbfa449 | Lars Kotthoff | 10 December 2017, 19:48:20 UTC | NEWS for #2099 | 10 December 2017, 19:48:20 UTC |
| ed18b6a | mb706 | 10 December 2017, 19:47:20 UTC | Automatically predict NA for rows w/ NAs and learners that don't support missings (#2099) * predict NA if learner doesn't support that * adding test * drop = FALSE * bugfix * using old prediction as fallback when all rows are NA * implementing @larskotthoff's suggestions | 10 December 2017, 19:47:19 UTC |
| c84b497 | Travis | 08 December 2017, 19:48:52 UTC | update auto-generated documentation [ci skip] | 08 December 2017, 19:48:52 UTC |
| b1e92be | Lars Kotthoff | 08 December 2017, 18:40:33 UTC | NEWS for 2081 | 08 December 2017, 18:40:33 UTC |
| f947f41 | Christian Scholbeck | 08 December 2017, 18:39:25 UTC | issue 1953 - added getOptPath function to tuneParams.R (#2081) * - added getOptPath function to tuneParams.R - it is a getter method for obtaining the optimization path from a tuning result * - changed naming according to mlr naming conventions * - changed the naming of the optimization path getter method * - added a unit test for the getTuneResultOptPath function - it tunes a minimal example svm and compares the output of the getter method with directly accessing the TuneResult object attribute - checks for equality of the returned opt.path as non-data frame and data frame output * - changed formatting according to lintr bots complaints * - deleted whitespaces * - deleted whitespace * - changed position of opening bracket * - fixed brackets | 08 December 2017, 18:39:25 UTC |
| 905421a | Lars Kotthoff | 07 December 2017, 02:48:33 UTC | remove spatial warning checks (#2090) | 07 December 2017, 02:48:33 UTC |
| f94e86d | Lars Kotthoff | 07 December 2017, 00:43:12 UTC | NEWS for #2080 | 07 December 2017, 00:43:12 UTC |
| 455f3e2 | Christian Scholbeck | 07 December 2017, 00:42:08 UTC | issue1460 - target distribution in taskDesc object (#2080) * added class.distribution table to classification task description; changed print.taskDesc function to print the attribute table and not calculate it anew * removed scholbeck_mlr.Rproj * - unittest does not use predefined iris task anymore but a newly created one * - reverse changes * - updated iris.task.rda in data folder, so that it contains the newly added target distribution table as an attribute in the task description * - changed the printing output of 'print.ClassifTask' - removed an empty line above the target distribution table - now prints the table via 'cat()' and adds a linebreak for the next line via '\n' in the following catf() statement * - changed position of comment * - updated tasks with target distribution as attribute * - changed data names to old names * - updated tasks | 07 December 2017, 00:42:08 UTC |
| 391c45a | Lars Kotthoff | 07 December 2017, 00:37:45 UTC | NEWS for #2085 | 07 December 2017, 00:37:45 UTC |
| a8b8c6d | alona-sydorova | 07 December 2017, 00:37:03 UTC | add adaboostm1 learner from rweka (#2085) * add adaboostm1 Learner from RWeka (implementation and test) * update style | 07 December 2017, 00:37:03 UTC |
| 596c954 | mb706 | 04 December 2017, 15:34:21 UTC | Removing superfluous print statement (#2084) | 04 December 2017, 15:34:21 UTC |
| 4858bc2 | Patrick Schratz | 03 December 2017, 12:25:56 UTC | fix data set loading errors | 03 December 2017, 12:25:56 UTC |
| 76c7235 | Tuan Le | 01 December 2017, 19:32:45 UTC | Fix Issue : Adding 'corr', 'nz' and 'nzv' method from caret::preProcess() (#2078) * #2050 add 'corr', 'nzv' and 'zv' method into preprocess * add 'corr', 'nzv' and 'zv' method into caret preprocWrapper (#2050) * (#2050) add 'corr', 'nzv' and 'zv' method into preprocess * add 'corr', 'nzv' and 'zv' method (#2050) * align all spaces for lintr * modified unit test to check output data frames after transformation * fixed lint | 01 December 2017, 19:32:45 UTC |
| 6868ecb | Tuan Le | 01 December 2017, 17:50:16 UTC | #2066 add 'prior' hyperparam into classif.lda learner (#2077) | 01 December 2017, 17:50:16 UTC |
| 6ebfdd7 | MiGraber | 01 December 2017, 16:45:05 UTC | added check for batchtools_issue 2072 (#2075) | 01 December 2017, 16:45:05 UTC |
| 780fcaa | Lars Kotthoff | 01 December 2017, 16:21:53 UTC | NEWS for #1891 | 01 December 2017, 16:21:53 UTC |
| fc28f58 | Travis | 30 November 2017, 22:56:37 UTC | update auto-generated documentation [ci skip] | 30 November 2017, 22:56:37 UTC |
| e865e49 | Patrick Schratz | 30 November 2017, 22:00:27 UTC | Add task attribute "is.spatial" and implement spatial partitioning method (#1891) * add basque data * implement spatial partitioning method based on kmeans * add test_classif_binomial_spatial * add binaryclass.spatial for tests * rename task attr "spatial" to "is.spatial" * remove x,y on condition, based on parent.call to work with nested resampling * cleanup * fix passing of spatial object * update test * update task datasets with 'is.spatial' attribute * fix assignment of 'is.spatial' in TaskDesc() * fix tests * add temporary testing files * lintr fix * fix examples * documentation * NEWS * man * travis check * add basque data * implement spatial partitioning method based on kmeans * add test_classif_binomial_spatial * add binaryclass.spatial for tests * rename task attr "spatial" to "is.spatial" * remove x,y on condition, based on parent.call to work with nested resampling * cleanup * fix passing of spatial object * update test * update task datasets with 'is.spatial' attribute * fix assignment of 'is.spatial' in TaskDesc() * fix tests * add temporary testing files * lintr fix * fix examples * documentation * NEWS * man * travis check * update test context * try to get travis build working again * SurvTasl rebase * travis * travis * delete testfiles * address requested changes * - checks for spatial task detection and respective warnings - update spatial task description * revert removal of empty line in travis.yml * fix warning message in getTaskData * add test for spatial partitioning * adapt ResampleInstances for test creation * fix tests * use viapply instead of sapply * restart travis * require bbmisc for warningf * add coords to doc of ResampleInstance * remove warning again - causes too much problems during tests. also, one warning in makeTask should be sufficient * fix wrong order of ResampleInstance.SpCV * account for trusty on travis * also account for 'train' and 'both' * add spatial attribute to new datasets * whitespace * update doc * add functionals item to task.desc$n.feat * update n.feat of sonar.task * add functionals to task.desc$n.feat for all datasets * update doc * restart travis * add functionals to agri.task * update warning messages * travis * travis * require both conditions to be TRUE when checking for coordinate names in data * update if conditions in makeTask* * new tests for if conditions in makeTask* | 30 November 2017, 22:00:27 UTC |
| d980377 | collinleiber | 29 November 2017, 18:56:32 UTC | Add different Ids for Multilabel Wrappers (#2071) * Changed the IDs from the different Multilabel Wrappers * Add test for issue 1700 * Use getLearnerId, getLearnerType and getLearnerPackages methods. Use stri_paste instead of paste | 29 November 2017, 18:56:32 UTC |
| f9ea122 | Alexander Engelhardt | 29 November 2017, 18:55:27 UTC | Fix for smote's wrong labeling of constant factors (#2063) * setting levels should only be done within the if, i.e. for factors * Write test for smote with constant factor features * Fix smote's wrong levels (issue 1951) * lintr style fixes | 29 November 2017, 18:55:27 UTC |
| 85cc831 | mb706 | 28 November 2017, 22:42:48 UTC | travis should notice build errors (#2070) * add check for missings in newdata * travis should notice build errors * Revert "add check for missings in newdata" This reverts commit f99e20d43106333953525ebb6e058502aa089097. | 28 November 2017, 22:42:48 UTC |
| 9319926 | smilesun | 28 November 2017, 14:33:35 UTC | fgam code change make more strict | 28 November 2017, 14:33:35 UTC |
| 652fa40 | ja-thomas | 28 November 2017, 12:22:40 UTC | fix batchmark for new batchtools version (#2067) * fix batchmark for new batchtools version * fix whitespaces | 28 November 2017, 12:22:40 UTC |
