« Changelog History


H2O v3.38.0.1 Release Notes

  • 🚀 Download at: http://h2o-release.s3.amazonaws.com/h2o/rel-zygmund/1/index.html

    🐛 Bug

    • 🛠 [PUBDEV-8811] - Fixed the sorting of h2o.make_leaderboard.
    • 🛠 [PUBDEV-8802] - Fixed H2O dependencies overriding Jetty implementation.
    • 🛠 [PUBDEV-8783] - Fixed Flow’s export Frame throwing an NPE because it doesn’t provide a file type.
    • 🛰 [PUBDEV-8765] - Fixed GLM ordinal generic metrics to provide missing information in the payload.
    • 🛠 [PUBDEV-8763] - Fixed “maxrsweep” NPE in ModelSelection thrown when the replacement step stopped too early.
    • ⚡️ [PUBDEV-8758] - Fixed “maxrsweep” replacement bug in ModelSelection by updating the implementation method.
    • 🐎 [PUBDEV-8742] - Fixed unnecessary transformations in the scikit-learn wrapper by using model performance API.
    • 🚀 [PUBDEV-8733] - Fixed upload of big files in Sparkling Water deployment.
    • 👍 [PUBDEV-8726] - Changed the error message that GLM does not support contributions.
    • 🛠 [PUBDEV-8722] - Fixed QuantilesGlobal histogram type failing in GBM when all columns were categorial.
    • ⚡️ [PUBDEV-8698] - Added support for MapR 6.2 to fix the error caused by updating the cluster.
    • 🛠 [PUBDEV-8693] - Fixed large file upload in Python.
    • 🖨 [PUBDEV-8670] - Fixed inability to stop print out of model information in Python.
    • 👀 [PUBDEV-8621] - Removed -seed variable hiding in GAM.
    • ⚡️ [PUBDEV-8561] - Updated h2o.upload_mojo to also work for POJO.
    • [PUBDEV-8220] - Added unsupported operation exception when trying to use SHAP summary plot when building DRF model with binomial_double_trees.
    • 🔨 [PUBDEV-7098] - Refactored the rendering logic in the Python client.
    • [PUBDEV-3524] - Added xval argument to h2o.confusionMatrix in R.

    👌 Improvement

    • 👍 [PUBDEV-8825] - Added support for calibrating an already trained model manually.
    • 👍 [PUBDEV-8801] - Added support for using Isotonic Regression for model calibration.
    • [PUBDEV-8795] - Added ability to S3A allowing it to share the built-in AWS credential providers.
    • [PUBDEV-8787] - Improved configure_s3_using_s3a allowing it to be usable in any deployment.
    • ⚡️ [PUBDEV-8727] - Updated train_segments function in R to be independent of camel casing in the algorithm name.
    • [PUBDEV-8721] - Improved runtime for QuantilesGlobal histogram by using exact split-points for low-cardinality columns.
    • [PUBDEV-8712] - Exposed the Sequential Walker for R/Python and added option to disable early stopping.
    • [PUBDEV-8692] - Cleaned up Key API by removing replicas.
    • 0️⃣ [PUBDEV-8315] - Cleaned up the default output after training a model.
    • [PUBDEV-8140] - Exposed calibrated probabilities in mojo_predict_pandas.

    🆕 New Feature

    • 🔧 [PUBDEV-8784] - Simplified the configuration of S3 for Frame exportation.
    • ✅ [PUBDEV-8725] - Added train_segments test for Isolation Forest.
    • [PUBDEV-8713] - Added ability to h2o.no_progress in R allowing it to accept expressions.
    • [PUBDEV-8685] - Implemented dispersion parameter estimation for GLM.
    • [PUBDEV-8679] - Added ability to export H2O Frame to a Parquet.
    • [PUBDEV-8589] - Added Pareto front plots to AutoML Explain.
    • [PUBDEV-8574] - Added “deviance” method to dispersion for calculating p-values.
    • [PUBDEV-8572] - Implemented variable inflation factors for GLM.
    • [PUBDEV-8470] - Implemented in-training checkpoints for GBM.
    • 👍 [PUBDEV-7634] - Implemented support for interactions to MOJO for CoxPH.
    • [PUBDEV-5280] - Added h2o.make_leaderboard function which scores and compares a set of models to AutoML.

    Task

    • [PUBDEV-8834] - Secured XGBoost connections in multinode environments.
    • [PUBDEV-8786] - Added missing added predictor and deleted predictor to the result frame and model summary of ModelSelection.
    • 🏗 [PUBDEV-8775] - Added support allowing you to force GLM to build a null model where the model only returns the coefficients for the intercept.
    • 👍 [PUBDEV-8774] - Added support allowing GLM gamma to fix the dispersion parameter to calculate p-values.
    • [PUBDEV-8703] - Implemented “maxr” speedup for Modelselection by introducting “maxrsweep”.
    • [PUBDEV-8683] - Implemented dispersion factor estimation using maximum likelihood for GLM gamma family.

    📄 Docs

    • 📚 [PUBDEV-8845] - Added documentation on Isotonic Regression.
    • [PUBDEV-8829] - Added variable inflation factors to GLM section of the user guide.
    • [PUBDEV-8827] - Added Tweedie dispersion parameter estimation to the GLM section of the user guide.
    • 🐎 [PUBDEV-8817] - Added confusion matrix calculation explanation to performance and prediction.
    • [PUBDEV-8806] - Added get_predictors_removed_per_step() and get_predictors_added_per_step() examples to ModelSelection.
    • [PUBDEV-8790] - Added use case section to the welcome page of the user guide.
    • [PUBDEV-8720] - Added MOJO import/export information to each algorithm page.
    • 🚀 [PUBDEV-8632] - Added major release blogs to user guide and moved change log to top of the sidebar.