Catboost

Bug Fixes and Other Changes
- Hotfix for critical bug in Python and R wrappers (issue 238)
- Added stratified data split in CV
- Fix `is_classification` check and CV for Logloss

Bug Fixes and Other Changes
- Fixed critical bugs in formula evaluation code (issue 236)
- Added scale_pos_weight parameter

Speedups
- 25% speedup of the model applier
- 43% speedup for training on large datasets.
- 15% speedup for `QueryRMSE` and calculation of querywise metrics.
- Large speedups when using binary categorical features.
- Significant (x200 on 5k trees and 50k lines dataset) speedup for plot and stage predict calculations in cmdline.
- Compilation time speedup.

Major Features And Improvements
- Industry fastest [applier implementation](https://tech.yandex.com/catboost/doc/dg/concepts/c-plus-plus-api-docpage/#c-plus-plus-api).
- Introducing new parameter [`boosting-type`](https://tech.yandex.com/catboost/doc/dg/concepts/python-reference_parameters-list-docpage/) to switch between standard boosting scheme and dynamic boosting, described in paper ["Dynamic boosting"](https://arxiv.org/abs/1706.09516).
- Adding new [bootstrap types](https://tech.yandex.com/catboost/doc/dg/concepts/python-reference_parameters-list-docpage/) `bootstrap_type`, `subsample`. Using `Bernoulli` bootstrap type with `subsample < 1` might increase the training speed.
- Better logging for cross-validation, added [parameter](https://tech.yandex.com/catboost/doc/dg/concepts/python-reference_cv-docpage/) `logging_level` and `metric_period` (should be set in training parameters) to cv.
- Added a separate `train` [function](https://tech.yandex.com/catboost/doc/dg/concepts/python-reference_train-docpage/) that receives the parameters and returns a trained model.
- Ranking mode `QueryRMSE` now supports default settings for dynamic boosting.
- R-package pre-build binaries are included into release.
- We added many synonyms to our parameter names, now it is more convenient to try CatBoost if you are used to some other library.

Bug Fixes and Other Changes
- Fix for CPU `QueryRMSE` with weights.
- Adding several missing parameters into wrappers.
- Fix for data split in querywise modes.
- Better logging.
- From this release we'll provide pre-build R-binaries
- More parallelisation.
- Memory usage improvements.
- And some other bug fixes.

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.

Major Features And Improvements
- We've made single document formula applier 4 times faster!
- `model.shrink` function added in [Python](https://tech.yandex.com/catboost/doc/dg/concepts/python-reference_catboost_shrink-docpage/) and R wrappers.
- Added new [training parameter](https://tech.yandex.com/catboost/doc/dg/concepts/python-reference_parameters-list-docpage/) `metric_period` that controls output frequency.
- Added new ranking [metric](https://tech.yandex.com/catboost/doc/dg/concepts/loss-functions-docpage/) `QueryAverage`.
- This version contains an easy way to implement new user metrics in C++. How-to example [is provided](https://github.com/catboost/tutorials/blob/master/custom_loss/custom_metric_tutorial.md).

Bug Fixes and Other Changes
- Stability improvements and bug fixes

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

Breaking Changes
Cmdline:
- Training parameter `gradient-iterations` renamed to `leaf-estimation-iterations`.
- `border` option removed. If you want to specify border for binary classification mode you need to specify it in the following way: `loss-function Logloss:Border=0.5`
- CTR parameters are changed:
- Removed `priors`, `per-feature-priors`, `ctr-binarization`;
- Added `simple-ctr`, `combintations-ctr`, `per-feature-ctr`;
More details will be published in our documentation.

Python:
- Training parameter `gradient_iterations` renamed to `leaf_estimation_iterations`.
- `border` option removed. If you want to specify border for binary classification mode you need to specify it in the following way: `loss_function='Logloss:Border=0.5'`
- CTR parameters are changed:
- Removed `priors`, `per_feature_priors`, `ctr_binarization`;
- Added `simple_ctr`, `combintations_ctr`, `per_feature_ctr`;
More details will be published in our documentation.

Major Features And Improvements
- In Python we added a new method `eval_metrics`: now it's possible for a given model to calculate specified metric values for each iteration on specified dataset.
- One command-line binary for CPU and GPU: in CatBoost you can switch between CPU and GPU training by changing single parameter value `task-type CPU` or `GPU` (task_type 'CPU', 'GPU' in python bindings). Windows build still contains two binaries.
- We have speed up the training up to 30% for datasets with a lot of objects.
- Up to 10% speed-up of GPU implementation on Pascal cards

Bug Fixes and Other Changes
- Stability improvements and bug fixes

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

Breaking Changes
FlatBuffers model format: new CatBoost versions wouldn’t break model compatibility anymore.

Major Features And Improvements
* Training speedups: we have speed up the training by 33%.
* Two new ranking modes are [available](https://tech.yandex.com/catboost/doc/dg/concepts/loss-functions-docpage/#ranking):
* `PairLogit` - pairwise comparison of objects from the input dataset. Algorithm maximises probability correctly reorder all dataset pairs.
* `QueryRMSE` - mix of regression and ranking. It’s trying to make best ranking for each dataset query by input labels.

Bug Fixes and Other Changes
* **We have fixed a bug that caused quality degradation when using weights < 1.**
* `Verbose` flag is now deprecated, please use `logging_level` instead. You could set the following levels: `Silent`, `Verbose`, `Info`, `Debug`.
* And some other bugs.

Thanks to our Contributors
This release contains contributions from: avidale, newbfg, KochetovNicolai and CatBoost team.

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