Skip to content

Latest commit

 

History

History
395 lines (387 loc) · 19.8 KB

CHANGELOG_detailed.md

File metadata and controls

395 lines (387 loc) · 19.8 KB

Changelog

0.4.1 (2017-11-06)

  • Untracking not so useful script. [Pradeep Reddy Raamana]

  • Test for arff usage. [Pradeep Reddy Raamana]

  • Improving font sizes in plots. [Pradeep Reddy Raamana]

  • Adding support for Weka's ARFF input datasets. [Pradeep Reddy Raamana]

  • Updating the flyer. [Pradeep Reddy Raamana]

  • Rest syntax fixes. [Pradeep Reddy Raamana]

  • Updating FAQ to reflect the latest implementation. [Pradeep Reddy Raamana]

  • Improvign the tests for visualization. [Pradeep Reddy Raamana]

  • Better handling of importance values for features never tested. [Pradeep Reddy Raamana]

  • Option to change classifier implemented!! [Pradeep Reddy Raamana]

  • Correcting stupid mistake. [Pradeep Reddy Raamana]

  • Minor bug fix in misclf freq calc. [Pradeep Reddy Raamana]

  • Getting the check of display check working on CI. [Pradeep Reddy Raamana]

  • Better corr with xtick in metric compare plot. [Pradeep Reddy Raamana]

  • Merge remote-tracking branch 'origin/master' [Pradeep Reddy Raamana]

  • Better checking of whether DISPLAY is available. [Pradeep Reddy Raamana]

  • Numbering the labels for accuracy plot for easy ref when n is large. [Pradeep Reddy Raamana]

  • Feat imp plots now handle never-selected features, display 95% CI only. [Pradeep Reddy Raamana]

  • Using non-interactive mode for pyplot only when DISPLAY is not set. [Pradeep Reddy Raamana]

  • Minor improvements to plots; version info to output logs. [Pradeep Reddy Raamana]

  • Using a different backend Agg for headless vis. [Pradeep Reddy Raamana]

  • Improving accuracy comparison plots, when legends are long; darker colors. [Pradeep Reddy Raamana]

  • Throwing in extra trees clf -- need to integrate and test. [Pradeep Reddy Raamana]

  • Fixing presentation of chance accuracy: training classes are stratified! [Pradeep Reddy Raamana]

  • Trying to implement API interface. [Pradeep Reddy Raamana]

  • New option to make vis from existing results [skip ci] [Pradeep Reddy Raamana]

  • Added the -version feature [skip ci] [Pradeep Reddy Raamana]

  • Logging datasets info common to all datasets for reference. [Pradeep Reddy Raamana]

  • Trying to disable parallelism within sklearn withOUT setting n_jobs; Option for NO grid search at all. [Pradeep Reddy Raamana]

  • Merge remote-tracking branch 'origin/master' [Pradeep Reddy Raamana]

  • Refactoring exhaustive search to be multiple levels; exposing in CLI. [Pradeep Reddy Raamana]

  • Add code of conduct [skip ci] [Pradeep Reddy Raamana]

  • Better docs. [Pradeep Reddy Raamana]

  • With pyradigms as input, no need to specify meta data separately; output folder is optional, writes results to $PWD/neuropredict. [Pradeep Reddy Raamana]

  • Link correction. [Pradeep Reddy Raamana]

  • Better docs. [Pradeep Reddy Raamana]

  • Makefile from cookiecutter. [Pradeep Reddy Raamana]

  • Not optimizing for min_impurity_decrease unless proven needed [skip ci] [Pradeep Reddy Raamana]

  • Falling back to sequential eval when #procs == 1. [Pradeep Reddy Raamana]

  • Much improved help text with raw formatting [skip ci] [Pradeep Reddy Raamana]

  • Choosing 50 reps for testing. [Pradeep Reddy Raamana]

  • More robust query of available cpu count. [Pradeep Reddy Raamana]

  • Preprocessing with robust scaling; option less exhaustive grid search; log flush. [Pradeep Reddy Raamana]

  • Setting RF param grid within RHsT itself. [Pradeep Reddy Raamana]

  • Boarding grid search scope for RF; more options for feat sel size. [Pradeep Reddy Raamana]

  • Merge branch 'parallel_rhst' [Pradeep Reddy Raamana]

  • Clean up unused; rename run_cli to cli. [Pradeep Reddy Raamana]

  • Merge pull request #16 from raamana/parallel_rhst. [Pradeep Reddy Raamana]

    Parallelizing CV repetitions

  • Adding auto versioning via versioneer (#15) [Kevin Le]

  • Finishing the parameter spec for the API [skip ci] [Pradeep Reddy Raamana]

  • Auto versioning. [Pradeep Reddy Raamana]

  • Update CONTRIBUTING.md. [Pradeep Reddy Raamana]

0.3.1 (2017-09-29)

  • Dropping support for python 2.7 :( [Pradeep Reddy Raamana]

  • Improving the chance accuracy tests. [Pradeep Reddy Raamana]

  • Finalizing option to choose a subgroup to process. [Pradeep Reddy Raamana]

  • Implement subgroups option. [Pradeep Reddy Raamana]

  • Option to choose a subgroup to process [skip ci] [Pradeep Reddy Raamana]

  • Fixing chance accuracy calculations and annotating it. [Pradeep Reddy Raamana]

  • New flag to specify number of CPUs [skip ci] [Pradeep Reddy Raamana]

  • Better gathering of feature importances [skip ci] [Pradeep Reddy Raamana]

  • Returning & gathering misclassification frequencies explicitly [skip ci] [Pradeep Reddy Raamana]

  • Rm unnecessary files [skip ci] [Pradeep Reddy Raamana]

  • Calculation of chance accuracy based on class imbalance [skip ci] [Pradeep Reddy Raamana]

  • Basic skeleton to gather results from parallel CV runs [skip ci] [Pradeep Reddy Raamana]

  • Tiny script generate visualizations from existing results. [Pradeep Reddy Raamana]

  • Basic skeletion for exporting first before visualization [ skip ci] [Pradeep Reddy Raamana]

    • parallelizing main loop over repetitions w/ shared data: to test [skip ci] - also saving results from each iteration independently - ignoring some warnings known to be bugs. [Pradeep Reddy Raamana]

  • To be tested init for parallel implementation. [Pradeep Reddy Raamana]

  • Adapting the new shape of confuison matrix container. [Pradeep Reddy Raamana]

  • Major refactoring of rhst module to streamline it via smaller methods. [Pradeep Reddy Raamana]

  • Parallelizign the main repetition loop in rhst; reformat code; [Pradeep Reddy Raamana]

  • Matching n_jobs with predispatch in gridsearch! [Pradeep Reddy Raamana]

  • Converting all asserts to if not raise. [Pradeep Reddy Raamana]

  • Parallel grid search with 4 cpus; minor changes to output and notes; [Pradeep Reddy Raamana]

  • Ignoring NaNs in computing medians, better outputs. [Pradeep Reddy Raamana]

  • Merge remote-tracking branch 'origin/master' [Pradeep Reddy Raamana]

  • Making the test dataset really small to speed up testing. [Pradeep Reddy Raamana]

  • Fixing # values to unpack in test. [Pradeep Reddy Raamana]

  • Reducing grid search time with smaller grid, efficient feat selectors [skip ci] [Pradeep Reddy Raamana]

  • Brute forcing module discovery: installing in devel mode on CI. [Pradeep Reddy Raamana]

  • Merge remote-tracking branch 'origin/master' [Pradeep Reddy Raamana]

  • Theme override try from mac. [Pradeep Reddy Raamana]

  • Adding static html files for gh-pages [skip ci] [Pradeep Reddy Raamana]

  • Getting sklearn pipeline and gridsearch combo to work. [Pradeep Reddy Raamana]

  • No argparse table wrap [skip ci] [Pradeep Reddy Raamana]

  • More docs [skip ci] [Pradeep Reddy Raamana]

  • Minor bug fixes and code cleanup. [Pradeep Reddy Raamana]

  • Falling back to test rudimentary RF and local grid search [skip ci] [Pradeep Reddy Raamana]

  • Ensuring each subgroup has atleast two unique classes; better docs;[skip ci] [Pradeep Reddy Raamana]

  • Better docs. [Pradeep Reddy Raamana]

  • Removing dead code [skip ci] [Pradeep Reddy Raamana]

  • Better reorg of input handling [skip ci] [Pradeep Reddy Raamana]

  • Fixing mistaken deletion [skip ci] [Pradeep Reddy Raamana]

  • Initial implementation of enabling generic selection of sklearn pipelines [skip ci] [Pradeep Reddy Raamana]

  • Treating the best params as a dict to enable use of models other than RF [skip ci] [Pradeep Reddy Raamana]

  • Finalizing grid search implementation [skip ci] [Pradeep Reddy Raamana]

  • Additional checks on feat sel size calc [skip ci] [Pradeep Reddy Raamana]

  • New method refactoring instantiation of pipeline & param grid [skip ci] [Pradeep Reddy Raamana]

  • Initial implementation of Pipeline mechanism: MI top K + RF; [skip ci] changes to dim red size calculation to make it slightly larger; [Pradeep Reddy Raamana]

  • Merge remote-tracking branch 'origin/master' [Pradeep Reddy Raamana]

  • Changing default parameters and steps a bit to speedup [skip ci] [Pradeep Reddy Raamana]

  • Trying out GridSearchCV for optimization of random forests - to be tested [skip ci] [Pradeep Reddy Raamana]

  • Quick skeleton for API access; better API docs [skip ci] [Pradeep Reddy Raamana]

  • Better non-occluded legends for cobweb misclf plots [skip ci] [Pradeep Reddy Raamana]

  • Closing figures as they are done exporting [skip ci] [Pradeep Reddy Raamana]

  • Merge remote-tracking branch 'origin/master' [Pradeep Reddy Raamana]

  • Set theme jekyll-theme-dinky. [Pradeep Reddy Raamana]

  • Ensuring the combined name doesnt get longer than allowed [skip ci] [Pradeep Reddy Raamana]

  • Improved docs [skip ci] [Pradeep Reddy Raamana]

  • Ensuring 1/num_classes appears on metric distribution for better presentation [skip ci] [Pradeep Reddy Raamana]

  • Test setup fixes for 2.7. [Pradeep Reddy Raamana]

  • New option to set size for feature selection; code/docs clean up. [Pradeep Reddy Raamana]

  • Module neuropredict removed, renamed to run_workflow.py. [Pradeep Reddy Raamana]

  • Merge remote-tracking branch 'origin/master' [Pradeep Reddy Raamana]

    Conflicts:

    neuropredict/main.py

  • Few renamings. [Pradeep Reddy Raamana]

  • Showing importance for only the top 25 features; python 3 support [skip ci] [Pradeep Reddy Raamana]

  • Major pkg reord to minimize confusion and circular imports; python 3 support tweaking. [Pradeep Reddy Raamana]

  • Option to input pyradigm as input; code cleanup. [Pradeep Reddy Raamana]

  • Better org for subgroup parsing. [Pradeep Reddy Raamana]

  • Not messing with css overrides - no text wrapping for argparse [skip ci] [Pradeep Reddy Raamana]

  • Adding the missed css file. [Pradeep Reddy Raamana]

  • Requirements file for docs [skip ci] [Pradeep Reddy Raamana]

  • Docs organized [skip ci] [Pradeep Reddy Raamana]

  • First try at docs [skip ci] [Pradeep Reddy Raamana]

  • Merge remote-tracking branch 'origin/master' [Pradeep Reddy Raamana]

  • Adding version badges [skip ci] [Pradeep Reddy Raamana]

  • Badges! [skip ci] [Pradeep Reddy Raamana]

  • Version info [skip ci] [Pradeep Reddy Raamana]

  • Separating parser for docs; shorthand for path handlers. [Pradeep Reddy Raamana]

  • Fewer samples, lower dim with more CV reps! parallelizing travis builds; [Pradeep Reddy Raamana]

  • Increasing timeout for CI and reducing dimensionality for faster CV runs. [Pradeep Reddy Raamana]

  • Omitting python 3.2-3.4. [Pradeep Reddy Raamana]

  • Trusting pytest/travis for import mechanism of 3.6 over PyCharm. [Pradeep Reddy Raamana]

  • Tests succeed in python 2.7/3.6 according to PyCharm - lets see about CI. [Pradeep Reddy Raamana]

  • Merge remote-tracking branch 'origin/master' [Pradeep Reddy Raamana]

  • Better description refering to subplots [skip ci] [Pradeep Reddy Raamana]

  • Markdown typo [skip ci] [Pradeep Reddy Raamana]

  • Answering why neuropredict [skip ci] [Pradeep Reddy Raamana]

  • Fixing a minor bug in warning message construction [skip ci] [Pradeep Reddy Raamana]

  • First steps towards python 3 compatibility; cleanup imports [skip ci] [Pradeep Reddy Raamana]

  • Confusion matrix title now reflects the feature set name. [Pradeep Reddy Raamana]

  • Updating FAQ to match the latest version. [Pradeep Reddy Raamana]

  • Adding link to large ML comparison study. [Pradeep Reddy Raamana]

  • Setting up CI. [Pradeep Reddy Raamana]

  • Merge remote-tracking branch 'origin/master' [Pradeep Reddy Raamana]

  • Correcting dependencies [skip ci] [Pradeep Reddy Raamana]

  • Update usage. [Pradeep Reddy Raamana]

  • Add Beerpay's badge. [Pradeep Reddy Raamana]

  • Update CONTRIBUTING.md. [Pradeep Reddy Raamana]

  • Update CONTRIBUTING.md. [Pradeep Reddy Raamana]

  • Adding the graphical slide on role and where it fits in the predictive analysis workflow. [Pradeep Reddy Raamana]

  • Improving the setup and requirements. [Pradeep Reddy Raamana]

  • Example datasets. [Pradeep Reddy Raamana]

  • Csv reading bug fix; adding setuptools to requirements. [Pradeep Reddy Raamana]

    • bug fix: when features.txt contains a single number, get method returns a 0-d scalar for which len() is not defined. Fixed it to return an array always. - feature addition: new visualization module to generate bar plots for comparison of misclassification rates for binary experiments - better handling of exceptions always throwing tracebacks - updated max_dimensionality_to_avoid_curseofdimensionality to not request more features than available. [Pradeep Reddy Raamana]

  • Adding slide on where np fits in the predictive analysis workflow. [Pradeep Reddy Raamana]

  • Adding link to overview slides. [Pradeep Reddy Raamana]

    • turning the exporting and reading in of results via a dictionary and not sure if this would solve the problem of having a fixed variable order. [Pradeep Reddy Raamana]

  • Create CONTRIBUTING.md. [Pradeep Reddy Raamana]

  • Italicizing questions for clarity. [Pradeep Reddy Raamana]

  • Delete freesurfer_features.py. [Pradeep Reddy Raamana]

    • new parallel coordinate plot for 2-class misclf rate visualization to better align with multi-class cobweb plot. [Pradeep Reddy Raamana]
    • bug fix: when features.txt contains a single number, get method returns a 0-d scalar for which len() is not defined. Fixed it to return an array always. - feature addition: new visualization module to generate bar plots for comparison of misclassification rates for binary experiments - better handling of exceptions: always throwing tracebacks - updated max_dimensionality_to_avoid_curseofdimensionality to not request more features than available. [Pradeep Reddy Raamana]

  • Removing build and egg-info from version tracking. [Pradeep Reddy Raamana]

    • additional assertion to ensure there are no duplicate sample ids in dataset - minor other fixes. [Pradeep Reddy Raamana]
    • feature addition: option to specify multiple subgroups in a N>2 multi-class experiment. [Pradeep Reddy Raamana]

  • Merge remote-tracking branch 'origin/master' [Pradeep Reddy Raamana]

  • FAQ arbitrary set of custom features. [Pradeep Reddy Raamana]

    • feature addition: support for numpy file format .npy as a data matrix. [Pradeep Reddy Raamana]
    • feature addition: implementation of reader for a data matrix in file. [Pradeep Reddy Raamana]

  • Removing empty old config file. [Pradeep Reddy Raamana]

    • separating the method misclassification percentage calculation - also saving the list of subjects that were selected for testing - including them for exporting results - removing unneeded parameters and cleanup. [Pradeep Reddy Raamana]
    • feature addition: allowing an arbitrary combination of multiple user-defined features and Freesurfer features. [Pradeep Reddy Raamana]
    • Comment on ROC curves. [Pradeep Reddy Raamana]
    • fixing the clipping of metric distribution at the bottom. [Pradeep Reddy Raamana]

  • Merge remote-tracking branch 'origin/master' [Pradeep Reddy Raamana]

  • FAQ on covariates. [Pradeep Reddy Raamana]

    • user prompt even if features for 1 subject could not be read - checking for the availability of x-server - more try/except to make the context more precise - correct name for scikit learn in requirements. [Pradeep Reddy Raamana]
    • fixing the entry point script - updating the readme with install troubleshooting. [Pradeep Reddy Raamana]
    • removing console script option in setup.py - slight cleanup - saving positive class specified to disk. [Pradeep Reddy Raamana]

  • Add install instructions and reorg. [Pradeep Reddy Raamana]

  • New file created as a reminder to implement nemenyi test. [Pradeep Reddy Raamana]

  • More generic and efificient code for minor things. [Pradeep Reddy Raamana]

  • Cleanup. [Pradeep Reddy Raamana]

  • Merge remote-tracking branch 'origin/master' [Pradeep Reddy Raamana]

  • Update FAQ.md. [Pradeep Reddy Raamana]

  • Merge remote-tracking branch 'origin/master' [Pradeep Reddy Raamana]

  • Update FAQ.md. [Pradeep Reddy Raamana]

  • New test to ensure chance accuracy on random datasets. [Pradeep Reddy Raamana]

  • Better reorg of the main entry module. [Pradeep Reddy Raamana]

  • Merge remote-tracking branch 'origin/master' [Pradeep Reddy Raamana]

  • Create FAQ.md. [Pradeep Reddy Raamana]

  • Update README.md. [Pradeep Reddy Raamana]

  • Initial trial for unit test for chance classifier. [Pradeep Reddy Raamana]

  • Fixing the export of confusion matrix. [Pradeep Reddy Raamana]

    • new module to export the different results to per-dataset CSVs - saving method names along with the results. [Pradeep Reddy Raamana]
    • feature names are integrated into the work flow (esp. feat imporatance) - freesurfer feature extraction methods return feature names/labels - tidying up the feature imp plot to horizontal bar plot. [Pradeep Reddy Raamana]

  • Merge remote-tracking branch 'origin/master' [Pradeep Reddy Raamana]

  • Flag names update. [Pradeep Reddy Raamana]

  • Improved misclassification histogram plots. [Pradeep Reddy Raamana]

  • Cleaner/more accurate names for functions. [Pradeep Reddy Raamana]

    • rough implementation of test for random dataset test. [Pradeep Reddy Raamana]
    • choosing the get method correctly for custom user input - cosmetics. [Pradeep Reddy Raamana]

  • Adding table of contents. [Pradeep Reddy Raamana]

  • Documentation for custom input features. [Pradeep Reddy Raamana]

  • Update README.md. [Pradeep Reddy Raamana]

  • Update README.md. [Pradeep Reddy Raamana]

  • Update README.md. [Pradeep Reddy Raamana]

  • Update README.md. [Pradeep Reddy Raamana]

  • Update README.md. [Pradeep Reddy Raamana]

  • Composite flyer for promotion. [Pradeep Reddy Raamana]

  • Fixing and organizing colormaps. [Pradeep Reddy Raamana]

  • Update README.md. [Pradeep Reddy Raamana]

  • Output images to be included in README. [Pradeep Reddy Raamana]

  • Updated README. [Pradeep Reddy Raamana]

  • Merge remote-tracking branch 'origin/master' [Pradeep Reddy Raamana]

  • Ignoring headerlines in the metadata file. [Pradeep Reddy Raamana]

    • two new flags to allow the user to choose training perc and num reps
    • new method to ensure reloaded dataset matches the meta data spec - implementation and integration of radar plots for comparison of pair- wise misclassification rates. [Pradeep Reddy Raamana]
    • positive class flag to enable AUC calculation - minor changes elsewhere. [Pradeep Reddy Raamana]
    • new reader for aseg stats whole brain using numpy.fromregex - minor changes elsewhere. [Pradeep Reddy Raamana]

  • Minor changes. [Pradeep Reddy Raamana]

    • implementation & integration of subcortical aseg stats - captured all the constant values to a config file. [Pradeep Reddy Raamana]
    • implementation & integration of module to summarize misclassified subjects - warning if features for too many subjects cant be read. [Pradeep Reddy Raamana]
    • Implemented and integrated the variable importance plots - Removed fixed seeding of random forest to make it truly random. [Pradeep Reddy Raamana]
    • Fixing a miscalculation in conf mat calculation and improving the presentation. [Pradeep Reddy Raamana]
    • implemented and integrated the accuracy and confusion matrix visualization - Tried to preserve label order in CM across different RHsT runs [TODO test] [Pradeep Reddy Raamana]

  • Reorg of the structure for packaging implementation of whole-brain seg volume reading basic integration test with fsvolumes reading and RHsT. [Pradeep Reddy Raamana]

  • Packaging setup and registration to PyPi. [Pradeep Reddy Raamana]

  • Renaming to neuropredict, improving the README Implementing Counters for misclassified sample ids. [Pradeep Reddy Raamana]

  • Preliminary implementation of RHsT and RF are done basic integration tests between RHsT and RF are done methods to save and load results done. [Pradeep Reddy Raamana]

  • Almost complete implementation of random forest optimization routine. TODO need to ensure class order is maintained over multiple trials of RHsT. [Pradeep Reddy Raamana]

  • Improving the flow in RHsT. [Pradeep Reddy Raamana]

  • Initial implementation of RHsT. [Pradeep Reddy Raamana]

    • Enabled methods to allow input of user-defined features. - Updated the readme with better usage notes. [Pradeep Reddy Raamana]

  • Initial barebones skeleton. [Pradeep Reddy Raamana]

  • Initial commit. [Pradeep Reddy Raamana]