Version 0.12.0

• Fixed crash with kappa when given two sets of ratings that are both
  missing an intermediate value (e.g., [1, 2, 4]).
• Added summarize_results script for creating a nice summary TSV file
  from a list of JSON results files.
• Summary files for ablation studies now have an extra column that says
  which feature was removed.

Version 0.11.0

• Added initial version of skll_convert script for converting between
  .jsonlines, .megam, and .tsv data file formats.
• Fixed bug in _megam_dict_iter where labels for instances with all zero
  features were being incorrectly set to None.
• Fixed bug in _tsv_dict_iter where features with zero values were being
  retained with values set as '0' instead of being removed completely. This
  caused DictVectorizer to create extra features, so results may
  change a little bit if you were using .tsv files.
• Fixed crash with predict and train_only modes when running on the grid.
• No longer use process pools to load files if
  SKLL_MAX_CONCURRENT_PROCESSES is 1.
• Added more informative error message when trying to load a file without
  any features.

Version 0.10.1

• Made processes non-daemonic to fix pool.map issue with running
  multiple configurations files at the same time with run_experiment.

Version 0.10.0

• run_experiment can now take multiple configuration files.
• Fixed issue where model parameters and scores were missing in evaluate
  mode

Version 0.9.17

• Added skll.data.convert_examples function to convert a list
  dictionaries to an ExamplesTuple.
• Added a new optional field to configuration file, ids_to_floats, to
  help save memory if you have a massive number of instances with numeric
  IDs.
• Replaced use_dense_features and scale_features options with
  feature_scaling. See the
  run_experiment documentation
  for details.

Version 0.9.16

• Fixed summary output for ablation experiments. Previously summary files
  would not include all results.
• Added ablation unit tests.
• Fixed issue with generating PDF documentation.

Version 0.9.15

• Added two new required fields to the configuration file format under the
  General heading: experiment_name and task. See the
  run_experiment documentation <http://skll.readthedocs.org/en/latest/run_experiment.html#creating-configuration-files>__
  for details.
• Fixed an issue where the "loading..." message was never being printed when
  loading data files.
• Fixed a bug where keyword arguments were being ignored for metrics when
  calculating final scores for a tuned model. This means that previous
  reported results may be wrong for tuning metrics that use keywords
  arguments: f1_score_micro, f1_score_macro,
  linear_weighted_kappa, and quadratic_weighted_kappa.
• Now try to convert IDs to floats if they look like them to save
  memory for very large files.
• kappa now supports negative ratings.
• Fixed a crash when specifing grid_search_jobs and pre-specified folds.

Version 0.9.14

• Hotfix to fix issue where grid_search_jobs setting was being overriden
  by grid_search_folds.

Version 0.9.13

• Added skll.data.write_feature_file (also available as
  skll.write_feature_file) to simplify outputting .jsonlines, .megam, and
  .tsv files.
• Added more unit tests for handling .megam and .tsv files.
• Fixed a bug that caused a crash when using gridmap.
• grid_search_jobs now sets both n_jobs and pre_dispatch for
  GridSearchCV under the hood. This prevents a potential memory issue when
  dealing with large datasets and learners that cannot handle sparse data.
• Changed logging format when using run_experiment to be a little more
  readable.

Version 0.9.12

• Fixed serious issue where merging feature sets was not working correctly.
  All experiments conducted using feature set merging (i.e., where you
  specified a list of feature files and had them merged into one set for
  training/testing) should be considered invalid. In general, your
  results should previously have been poor and now should be much better.
• Added more verbose regression output including descriptive statistics
  and Pearson correlation.