Release 0.11.1

Changes:

• Accelerated formula evaluation by ~15%
• Improved model application interface
• Improved compilation time for building GPU version
• Better handling of stray commas in list arguments
• Added a benchmark that employs Rossman Store Sales dataset to compare quality of GBDT packages
• Added references to Catboost papers in R-package CITATION file (issue #488)
• Fixed a build issue in compilation for GPU
• Fixed a bug in model applicator
• Fixed model conversion (issue #533)
• Returned pre 0.11 behaviour for best_score_ and evals_result_ (issue #539)
• Make valid .dist-info/RECORD in python wheel (issue #534)

Changes:

• Changed default border count for float feature binarization to 254 on CPU to achieve better quality
• Fixed random seed to 0 by default
• Support model with more than 254 feature borders or one hot values when doing predictions
• Added model summation support in python: use catboost.sum_models() to sum models with provided weights.
• Added json model tutorial model_export_as_json_tutorial.ipynb

Changes:

• Added EvalResult output after GPU catboost training

• Supported prediction type option on GPU

• Added get_evals_result() method and evals_result_ property to model in python wrapper to allow user access metric values

• Supported string labels for GPU training in cmdline mode

• Many improvements in JNI wrapper

• Updated NDCG metric: speeded up and added NDCG with exponentiation in numerator as a new NDCG mode

• CatBoost doesn't drop unused features from model after training

• Write training finish time and catboost build info to model metadata

• Fix automatic pairs generation for GPU PairLogitPairwise target

Main changes:

• Fixed Python 3 support in catboost.FeaturesData
• 40% speedup QuerySoftMax CPU training

Bugfixes

• Fixed #403 bug in cuda train submodule (training crashed without evaluation set)
• Fixed exception propagation on pool parsing stage
• Add support of string GroupId and SubgroupId in python-package
• Print real class names instead of their labels in eval output

Breaking Changes

• We removed calc_feature_importance parameter from Python and R. Now feature importance calculation is almost free, so we always calculate feature importances. Previously you could disable it if it was slowing down your training.
• We removed Doc type for feature importances. Use Shap instead.
• We moved thread_count parameter in Python get_feature_importance method to the end.

Ranking

In this release we added several very powerfull ranking objectives:

• PairLogitPairwise
• YetiRankPairwise
• QueryCrossEntropy (GPU only)

Other ranking improvements:

• We have made improvements to our existing ranking objectives QuerySoftMax and PairLogit.
• We have added group weights support.

Accuracy improvements

• Improvement for datasets with weights
• Now we automatically calculate a good learning rate for you in the start of training, you don't have to specify it. After the training has finished, you can look on the training curve on evaluation dataset and make ajustments to the selected learning rate, but it will already be a good value.

Speedups:

• Several speedups for GPU training.
• 1.5x speedup for applying the model.
• Speed up multi classificaton training.
• 2x speedup for AUC calculation in eval_metrics.
• Several speedups for eval_metrics for other metrics.
• 100x speed up for Shap values calculation.
• Speedup for feature importance calculation. It used to be a bottleneck for GPU training previously, now it's not.
• We added possibility to not calculate metric on train dataset using MetricName:hint=skip_train~false (it might speed up your training if metric calculation is a bottle neck, for example, if you calculate many metrics or if you calculate metrics on GPU).
• We added possibility to calculate metrics only periodically, not on all iterations. Use metric_period for that. (previously it only disabled verbose output on each iteration).
• Now we disable by default calculation of expensive metrics on train dataset. We don't calculate AUC and PFound metrics on train dataset by default. You can also disable calculation of other metrics on train dataset using MetricName:hints=skip_train~true. If you want to calculate AUC or PFound on train dataset you can use MetricName:hints=skip_train~false.
• Now if you want to calculate metrics using eval_metrics or during training you can use metric_period to skip some iterations. It will speed up eval_metrics and it might speed up training, especially GPU training. Note that the most expensive metric calculation is AUC calculation, for this metric and large datasets it makes sense to use metric_period. If you only want to see less verbose output, and still want to see metric values on every iteration written in file, you can use verbose=n parameter
• Parallelization of calculation of most of the metrics during training

Improved GPU experience

• It is possible now to calculate and visualise custom_metric during training on GPU. Now you can use our Jupyter visualization, CatBoost viewer or TensorBoard the same way you used it for CPU training. It might be a bottleneck, so if it slows down your training use metric_period=something and MetricName:hint=skip_train~false
• We switched to CUDA 9.1. Starting from this release CUDA 8.0 will not be supported
• Support for external borders on GPU for cmdline

Improved tools for model analysis

• We added support of feature combinations to our Shap values implementation.
• Added Shap values for MultiClass and added an example of it's usage to our Shap tutorial.
• Added prettified parameter to get_feature_importance(). With prettified=True the function will return list of features with names sorted in descending order by their importance.
• Improved interfaces for eval-feature functionality
• Shap values support in R-package

New features

• It is possible now to save any metainformation to the model.
• Empty values support
• Better support of sklearn
• feature_names_ for CatBoost class
• Added silent parameter
• Better stdout
• Better diagnostic for invalid inputs
• Better documentation
• Added a flag to allow constant labels

New metrics

We added many new metrics that can be used for visualization, overfitting detection, selecting of best iteration of training or for cross-validation:

• BierScore
• HingeLoss
• HammingLoss
• ZeroOneLoss
• MSLE
• MAE
• BalancedAccuracy
• BalancedErrorRate
• Kappa
• Wkappa
• QueryCrossEntropy
• NDCG

New ways to apply the model

• Saving model as C++ code
• Saving model with categorical features as Python code

New ways to build the code

Added make files for binary with CUDA and for Python package

Tutorials

We created a new repo with tutorials, now you don't have to clone the whole catboost repo to run Jupyter notebook with a tutorial.

Bugfixes

We have also a set of bugfixes and we are gratefull to everyone who has filled a bugreport, helping us making the library better.

Thanks to our Contributors

This release contains contributions from CatBoost team. We want to especially mention @pukhlyakova who implemented lots of useful metrics.

As usual we are grateful to all who filed issues or helped resolve them, asked and answered questions.

Bug Fixes and Other Changes

• New model method get_cat_feature_indices() in Python wrapper.
• Minor fixes and stability improvements.

Breaking changes

• We fixed bug in CatBoost. Pool initialization from numpy.array and pandas.dataframe with string values that can cause slight inconsistence while using trained model from older versions. Around 1% of cat feature hashes were treated incorrectly. If you expirience quality drop after update you should consider retraining your model.

Major Features And Improvements

• Algorithm for finding most influential training samples for a given object from the 'Finding Influential Training Samples for Gradient Boosted Decision Trees' paper is implemented. This mode for every object from input pool calculates scores for every object from train pool. A positive score means that the given train object has made a negative contribution to the given test object prediction. And vice versa for negative scores. The higher score modulo - the higher contribution. See get_object_importance model method in Python package and ostr mode in cli-version. Tutorial for Python is available here. More details and examples will be published in documentation soon.
• We have implemented new way of exploring feature importance - SHAP values from paper. This allows to understand which features are most influent for a given object. You can also get more insite about your model, see details in a tutorial.
• Save model as code functionality published. For now you could save model as Python code with categorical features and as C++ code w/o categorical features.

Bug Fixes and Other Changes

• Fix _catboost reinitialization issues #268 and #269.
• GPU parameter use_cpu_ram_for_cat_features renamed to gpu_cat_features_storage with posible values CpuPinnedMemory and GpuRam. Default is GpuRam.

Thanks to our Contributors

This release contains contributions from CatBoost team.

As usual we are grateful to all who filed issues or helped resolve them, asked and answered questions.

Major Features And Improvements

• GPU: New DocParallel mode for tasks without categorical features and with —max-ctr-complextiy 1. Provides best performance for pool with big number of documents.
• GPU: Distributed training on several GPU host via MPI. See instruction how to build binary here.
• GPU: Up to 30% learning speed-up for Maxwell and later GPUs with binarization level > 32

Bug Fixes and Other Changes

• Hotfixes for GPU version of python wrapper.

Major Features And Improvements

• Python wrapper: added methods to download datasets titanic and amazon, to make it easier to try the library (catboost.datasets).
• Python wrapper: added method to write column desctiption file (catboost.utils.create_cd).
• Made improvements to visualization.
• Support non-numeric values in GroupId column.
• Tutorials section updated.

Bug Fixes and Other Changes

• Fixed problems with eval_metrics (issue #285)
• Other fixes

Breaking changes

• Changed parameter order in train() function to be consistant with other GBDT libraries.
• use_best_model is set to True by default if eval_set labels are present.

Major Features And Improvements

• New ranking mode YetiRank optimizes NDGC and PFound.
• New visualisation for eval_metrics and cv in Jupyter notebook.
• Improved per document feature importance.
• Supported verbose=int: if verbose > 1, metric_period is set to this value.
• Supported type(eval_set) = list in python. Currently supporting only single eval_set.
• Binary classification leaf estimation defaults are changed for weighted datasets so that training converges for any weights.
• Add model_size_reg parameter to control model size. Fix ctr_leaf_count_limit parameter, also to control model size.
• Beta version of distributed CPU training with only float features support.
• Add subgroupId to Python/R-packages.
• Add groupwise metrics support in eval_metrics.

Thanks to our Contributors

This release contains contributions from CatBoost team.

We are grateful to all who filed issues or helped resolve them, asked and answered questions.

Breaking changes

• boosting_type parameter value Dynamic is renamed to Ordered.
• Data visualisation functionality in Jupyter Notebook requires ipywidgets 7.x+ now.
• query_id parameter renamed to group_id in Python and R wrappers.
• cv returns pandas.DataFrame by default if Pandas installed. See new parameter as_pandas.

Major Features And Improvements

• CatBoost build with make file. Now it’s possible to build command-line CPU version of CatBoost under Linux with make file.
• In column description column name Target is changed to Label. It will still work with previous name, but it is recommended to use the new one.
• eval-metrics mode added into cmdline version. Metrics can be calculated for a given dataset using a previously trained model.
• New classification metric CtrFactor is added.
• Load CatBoost model from memory. You can load your CatBoost model from file or initialize it from buffer in memory.
• Now you can run fit function using file with dataset: fit(train_path, eval_set=eval_path, column_description=cd_file). This will reduce memory consumption by up to two times.
• 12% speedup for training.

Bug Fixes and Other Changes

• JSON output data format is changed.
• Python whl binaries with CUDA 9.1 support for Linux OS published into the release assets.
• Added bootstrap_type parameter to CatBoostClassifier and Regressor (issue #263).

Thanks to our Contributors

This release contains contributions from newbfg and CatBoost team.

We are grateful to all who filed issues or helped resolve them, asked and answered questions.

Major Features And Improvements

• BETA version of distributed mulit-host GPU via MPI training
• Added possibility to import coreml model with oblivious trees. Makes possible to migrate pre-flatbuffers model (with float features only) to current format (issue #235)
• Added QuerySoftMax loss function

Bug Fixes and Other Changes

• Fixed GPU models bug on pools with both categorical and float features (issue #241)
• Use all available cores by default
• Default float features binarization method set to GreedyLogSum
• Fixed not querywise loss for pool with QueryId

Bug Fixes and Other Changes

• Hotfix for critical bug in Python and R wrappers (issue #238)
• Fix is_classification check and CV for Logloss (issue #237)
• Added stratified data split in CV

Bug Fixes and Other Changes

• Fixed critical bugs in formula evaluation code (issue #236)
• Added scale_pos_weight parameter

Breaking Changes

• R library interface significantly changed
• New model format: CatBoost v0.2 model binary not compatible with previous versions
• Cross-validation parameters changes: we changed overfitting detector parameters of CV in python so that it is same as those in training.
• CTR types: MeanValue => BinarizedTargetMeanValue

Major Features And Improvements

• Training speedups: we have speed up the training by 20-30%.
• Accuracy improvement with categoricals: we have changed computation of statistics for categorical features, which leads to better quality.
• New type of overfitting detector: Iter. This type of detector was requested by our users. So now you can also stop training by a simple criterion: if after a fixed number of iterations there is no improvement of your evaluation function.
• TensorBoard support: this is another way of looking on the graphs of different error functions both during training and after training has finished. To look at the metrics you need to provide train_dir when training your model and then run "tensorboard --logdir={train_dir}"
• Jupyter notebook improvements: for our Python library users that experiment with Jupyter notebooks, we have improved our visualisation tool. Now it is possible to save image of the graph. We also have changed scrolling behaviour so that it is more convenient to scroll the notebook.
• NaN features support: we also have added simple but effective way of dealing with NaN features. If you have some NaNs in the train set, they will be changed to a value that is less than the minimum value or greater than the maximum value in the dataset (this is configurable), so that it is guaranteed that they are in their own bin, and a split would separates NaN values from all other values. By default, no NaNs are allowed, so you need to use option nan_mode for that. When applying a model, NaNs will be treated in the same way for the features where NaN values were seen in train. It is not allowed to have NaN values in test if no NaNs in train for this feature were provided.
• Snapshotting: we have added snapshotting to our Python and R libraries. So if you think that something can happen with your training, for example machine can reboot, you can use snapshot_file parameter - this way after you restart your training it will start from the last completed iteration.
• R library tutorial: we have added tutorial.
• Logging customization: we have added allow_writing_files parameter. By default some files with logging and diagnostics are written on disc, but you can turn it off using by setting this flag to False.
• Multiclass mode improvements: we have added a new objective for multiclass mode - MultiClassOneVsAll. We also added class_names param - now you don't have to renumber your classes to be able to use multiclass. And we have added two new metrics for multiclass: TotalF1 and MCC metrics. You can use the metrics to look how its values are changing during training or to use overfitting detection or cutting the model by best value of a given metric.
• Any delimeters support: in addition to datasets in tsv format, CatBoost now supports files with any delimeters

Bug Fixes and Other Changes

• Stability improvements and bug fixes

Thanks to our Contributors

This release contains contributions from: grayskripko, hadjipantelis and CatBoost team.

We are grateful to all who filed issues or helped resolve them, asked and answered questions.