Skip to content
Permalink
Branch: master
Find file Copy path
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
366 lines (295 sloc) 33.1 KB

Change Log

v1.0.0.20191110 (2019-11-10)

Full Changelog

Implemented enhancements:

  • Enforce a code style #208

Fixed bugs:

  • Add CI for python 3.6 #259

Closed issues:

  • Fix common segfaults #260
  • TPOT segmentation fault #257
  • TypeError: can't pickle _thread.RLock objects when saving a pipe #253

Merged pull requests:

v2019.10.14 (2019-10-15)

Full Changelog

Implemented enhancements:

  • More pipeline diagnostics #239
  • add dependabot #236
  • Add .from_preset method to MatPipe #232
  • Change or add to digest to make it easier to read #221
  • Featurizer sets can be more easily rewritten as dataclasses #209
  • rm /tutorials and add to matminer_examples #205
  • Add warning when large numbers of samples imputed/handled #199

Fixed bugs:

  • Code docs need overhaul #244
  • Pipeline save/load with TPOT backend doesn't attrs in intuitive way #241
  • MatPipe.load should refuse to load from class instance #234
  • Automatminer save/load needs more robust test #231
  • Add warning for mismatched versions of automatminer on save/load #230
  • Fix requirements #229
  • Autofeaturizer caching needs to use matminer utils, not pd.json #226
  • initialize_logger has confusing arguments #204

Closed issues:

  • Reassign pipe logger #242
  • MatPipe save/load does not work on TPOTAdaptor pipelines #235
  • Add ability to MatPipe to suppress internal warnings #233
  • Make it easier to ignore entire columns but keep them in returned df #228
  • Benchdev needs a workflow for predicting properties #227
  • add function to matpipe to output pipeline as simple script? #224
  • VERSION FileNotFoundError on import #223
  • Add automatminer citation to all matbench datasets #218
  • Docs suck #216
  • Rewrite analytics to MatPipe #186

Merged pull requests:

  • serialize backend and test improvements #246 (ardunn)
  • refactor setting loggers #243 (janosh)
  • Add support for pipeline digest in JSON and YAML format #238 (janosh)

v2019.9.11 (2019-09-11)

Full Changelog

v2019.9.12 (2019-09-11)

Full Changelog

Closed issues:

  • benchdev needs to be updated with newest matbench v0.1 names #222
  • benchdev infrastructure changes #220

v2019.08.07_beta (2019-08-08)

Full Changelog

v2019.08.07_betaK (2019-08-08)

Full Changelog

Closed issues:

  • fix failing tests #215
  • remove target from predict? #214
  • Cannot rebuild docs? #213
  • Consider replacing XGBoost with Catboost #195
  • TPOT will, on occasion, randomly fail #181
  • Make an autokeras adaptor #147
  • Look at skater rule based models as a solution for small datasets #145

v2019.05.14_beta0 (2019-05-15)

Full Changelog

Closed issues:

  • Update xgboost version to 0.80 #210
  • featurization takes too much ram #206
  • setup.py imports automatminer #202
  • Include a (basic) neural network separate from NNAdaptor #197
  • Metaselector needs a rework #149

Merged pull requests:

v2019.03.27_beta (2019-03-27)

Full Changelog

Merged pull requests:

v2019.03.26b0 (2019-03-27)

Full Changelog

Closed issues:

  • matpipe benchmark does not work with StratifiedKFold #191
  • Change TPOT default optimization metric to MAE? #190
  • PCA fails if matrix is large and n_features > n_samples #183
  • Add "powerups" to presets #180
  • General problem: Featurization takes too long! #179
  • DataCleaner na_method is sloppy #178
  • Autofeaturizer logging is annoying #177
  • Autofeaturizer may run redundant conversions as many as 3 times #176
  • Add circleci test for 3.7 #175
  • Logger should append to existing logs, not overwrite it #174
  • Analytics tests should run whether or not they are on circleci #169
  • Real docs + more thorough example #167
  • Add option to control tree and correlation-based FeatureReducer params #162
  • Use MEGNet/CGCNN as backend? #156
  • Need more featurizers implemented in matminer/automatminer #143
  • Outlier detection as a preprocessing step #135
  • Look into FunctionFeaturizer #134
  • Analysis class needs to be beefed up with something actually useful #105
  • Analysis should produce summary and visualization as file #57

Merged pull requests:

v2019.02.03_beta (2019-02-03)

Full Changelog

v2019.02.02_beta (2019-02-02)

Full Changelog

Closed issues:

  • Nose ---> unittest #171
  • Fix benchmarking #170
  • Should add to PyPi #168
  • An adapter to run a single model #165
  • Add option to remove specific features #159
  • Analytics module needs tests #133

Merged pull requests:

  • Update codacy and circleCI configs #173 (utf)
  • Add optional to manually keep/remove features #172 (utf)

v2019.01.26_beta (2019-01-26)

Full Changelog

Closed issues:

  • MatPipe code needs revamp #166
  • Implement nested CV for pipeline benchmarking #163
  • CircleCI + package reqs needs update #150

v2019.01.25_beta (2019-01-26)

Full Changelog

Closed issues:

  • List-like test_spec behavior broken in benchmark function #161
  • Add random_state option to benchmark function #160
  • Autofeaturizer needs ability to use custom column names #148
  • Add jarvis to AllFeaturizers #144
  • Empty logger made anytime MatPipe imported #141
  • removing correlated features doesn't work for classification targets #140
  • Oxidation states guessed twice #138
  • Look into using NestedCV for automl, and whether it would be a good idea or not #130
  • Add a very simple example #108
  • Add another study comparison with matbench #65
  • Add a profiler to DataframeTransformer #56

Merged pull requests:

  • Added Examples Folder and MSE Example #158 (ADA110)
  • Analytics Module Tests #157 (ADA110)
  • (WIP) Custom Column Names #152 (ADA110)
  • Better handling of adding oxidation states to large structures. #142 (utf)

2018.12.11_beta (2018-12-12)

Full Changelog

Closed issues:

  • No module named 'automatminer.featurize' #146
  • metaselection needs error handling, or screening beforehand #139
  • Add logging warning level option to Matpipe object. #136
  • Add ability to ensemble top models in tpot #111
  • Make dataset test set #107
  • Consider adaptor classes for other backends #100
  • Using skater for analysis #95
  • Tpot defaults need investigation and modification #79
  • Add more featurizers to AllFeaturizers #60

Merged pull requests:

v2018.11.16-beta (2018-11-17)

Full Changelog

v2018.11.2-beta (2018-11-17)

Full Changelog

Closed issues:

  • MatPipe cannot be serialized #124
  • MatPipe needs tests #118
  • DataCleaner needs scaling #115
  • Logger problems #114
  • Remove temp fix of CompositionToOxidComposition with next matminer #113
  • Tpot tests need to be redone #109
  • Top level classes should be able to serialize all pipeline info to json #102
  • Top level class methods need work #97
  • Tests need coverage assessment #83
  • Dataset storage needs improvement #80

Merged pull requests:

  • update benchmark on matpipe and update test #129 (ardunn)
  • Add matpipe tests, digest, and ability to save and load pipes #127 (ardunn)
  • MatPipe improvements + tpot tests #126 (ardunn)
  • Refactor mslearn to use matminer for its dataset needs #125 (Doppe1g4nger)
  • remove conversion override and fix log typos #123 (ardunn)
  • quick update verbosity of adaptor #122 (ardunn)
  • big ol refactor + matpipe updates #121 (ardunn)
  • make metaselection part of AutoFeaturizer #120 (Qi-max)
  • Fix pipeline logger #119 (utf)

v2018.11.2-alpha (2018-11-02)

Closed issues:

  • Logger needed #103
  • Heuristic based featurizer selection #99
  • ++robustness and usefulness of featurizer sets #98
  • Convert preprocessing modele into 2 separate operations #96
  • Adding/converting .fit/.transform/.predict methods #92
  • Further package structure suggestions #88
  • Rename AutoML segment of pipeline to better reflect package use #87
  • Plan - WIP #85
  • TestAllFeaturizers will break whenever a new featurizer is added #82
  • test_tpot hard to maintain with tpot version update #73
  • Dimensionality reduction #61
  • matminer issue: MaximumPackingEfficiency error #59
  • set mpid as index if available in load_* functions #58
  • Castelli is missing structure or the doc is incorrect #54
  • Testing takes unacceptably long times #49
  • zhou gaps formula can't be converted to composition #38
  • Model Selection Methodology #33
  • normalize preprocess for the future use of pipeline #20
  • find a way to obtain feature_cols list and target_col easily #19
  • load_* functions should ensure all numeric columns #16
  • load_mp should return other quantities #15
  • all formula columns in load_* funcs should return Composition objects #14
  • What is MatbenchError? #11

Merged pull requests:

  • Adding top level class + bugfixes #117 (ardunn)
  • change logging default and show example #116 (ardunn)
  • a base to start from #112 (ardunn)
  • Matbench wide logger #110 (utf)
  • Improve metalearning for automatically filter featurizers #106 (Qi-max)
  • add cv docs + citations and implementors methods #104 (albalu)
  • add TreeBasedFeatureReduction + tests #101 (albalu)
  • Adding top level class skeleton and jarvis dataset #94 (ardunn)
  • Update dataset loading utilities to use new function #93 (Doppe1g4nger)
  • organize preprocess just a bit #91 (albalu)
  • Improved testing of AllFeaturizers class #90 (utf)
  • Update conversions to use conversion featurizers #89 (utf)
  • make subpackages and their importing consistent #86 (albalu)
  • consistent naming #84 (albalu)
  • Organize file structure, add a dataset + more #81 (ardunn)
  • Code cleanup on tpot_utils #78 (Doppe1g4nger)
  • Cleanup is_greater_better function #76 (Doppe1g4nger)
  • fix logger issue duplicated in notebooks #75 (albalu)
  • Add classifier/regressor config_dicts for customizing pipeline operators #74 (Qi-max)
  • Split + improve existing glass datasets and add a new dataset #72 (Qi-max)
  • Change default for max_na_frac, add notebook #71 (ardunn)
  • Fix issue with failing when no column #70 (ardunn)
  • Featurize changes #69 (ardunn)
  • Restructure project #68 (ardunn)
  • Organize featurize #67 (ardunn)
  • correct weights + multi-class #66 (albalu)
  • cleanup of the gap notebook #64 (albalu)
  • add notebook predicting experimental band gap dataset #63 (albalu)
  • a more complete example #62 (albalu)
  • featurize_column + template for tricky target #55 (albalu)
  • multiindex and relevant changes #53 (albalu)
  • matminer version #52 (albalu)
  • target_col -> target #51 (albalu)
  • prune_correlated_features #50 (albalu)
  • Add matminer flla dataset #48 (ardunn)
  • max_colnull=0.05 default #47 (albalu)
  • started feature_importance #46 (albalu)
  • Fixing boltztrap_mp and phonon datasets to have formula and structures #45 (ardunn)
  • Improve metafeature design and add unittests for current composition/structure metafeatures #44 (Qi-max)
  • Benchmark autosklearn #43 (Qi-max)
  • exclude featurize_formula that need oxidations #42 (albalu)
  • separate slow composition featurizers #41 (albalu)
  • fixed_expt_gap #40 (albalu)
  • fixed bad formulas in load_expt_gap #39 (albalu)
  • start ErrorAnalysis #37 (albalu)
  • start TpotAutoml #36 (albalu)
  • add steel_strength dataset and Heusler_magnetic dataset #35 (Qi-max)
  • add load_boltztrap_mp #34 (albalu)
  • small change in featurize + matbench tpot example #32 (albalu)
  • expand PreProcess a bit #31 (albalu)
  • Initial Design of the Composition/Structure/Missing_value-related Metafeatures for the Input Dataset #30 (Qi-max)
  • added tests for AllFeaturizers #29 (albalu)
  • additional featurizers and tests #28 (albalu)
  • start dos featurizers #27 (albalu)
  • the rest of structure featurizers #26 (albalu)
  • 1st attempt to make a complete collection of featurizers #25 (albalu)
  • Add metalearning and some thoughts on meta-feature design, strategy and recommendation #24 (Qi-max)
  • more organization for featurization #23 (albalu)
  • Make all load_* return dict Structures and string Compositions (as pretty_formula) #22 (ardunn)
  • improved metrics mapping in automl #21 (Qi-max)
  • fix drop None in Featurize init #18 (Qi-max)
  • added a class to perform auto machine learning #17 (Qi-max)
  • fix loading mp #13 (ardunn)
  • update doc in load #12 (ardunn)
  • Add 4 datasets, cleanup + make load_mp better #10 (ardunn)
  • add xlrd to reqs to support pd.read_excel #9 (albalu)
  • add tox.ini #8 (albalu)
  • add test_data to get testing started #7 (albalu)
  • featurize just featurizes no cleanup, etc #6 (albalu)
  • test PR direct from fork #5 (albalu)
  • get preprocess started #4 (albalu)
  • Doc updates and renaming #3 (ardunn)
  • moved packages inside matbench for expected behavior #2 (albalu)
  • Test PR #1 (ardunn)

* This Change Log was automatically generated by github_changelog_generator

You can’t perform that action at this time.