H2O v3.32.1.1 Release Notes
-
π Download at: http://h2o-release.s3.amazonaws.com/h2o/rel-zipf/1/index.html
π Bug
[PUBDEV-6356] - GBM histograms now ignore rows with NA responses. [PUBDEV-7606] - Variable Importances added to GLM Generic model. π [PUBDEV-7782] - Fixed the ArrayIndexOutOfBoundsException issue with GLM CV. π [PUBDEV-7825] - CoxPH performance no longer fails when a factor is used for the
event_column
. [PUBDEV-7841] - Existing frame no longer overwritten when data with the same query is loaded. π [PUBDEV-7909] - Fixed howgain
is calculated in XGBFI for GBM. [PUBDEV-7934] - Improved the error messages forsave_to_hive_table
. β [PUBDEV-7963] - Added missing argument βtestβ forh2o.explain_row()
. π¨ [PUBDEV-7979] - All trees now supported for XGBoost Print MOJO in Java. [PUBDEV-7987] - CoxPHprediction
no longer fails whenoffset_column
is specified. [PUBDEV-7998] - Added keys for Individual Conditional Expectation (ICE) plot in H2OExplanation class. [PUBDEV-8013] -model@model$parameters$x
now reports actual feature names instead ofnames
. [PUBDEV-8016] -h2o.explain
no longer errors when AutoML object is trained with afold_column
. π [PUBDEV-8046] - Fixed issues with pythonβs explanation plots not displaying fully.π New Feature
[PUBDEV-7706] - Ignored columns that are actually used for model training are unignored and no longer prevent model training to start in Flow. [PUBDEV-7735] - Added baseline hazard function estimate to CoxPH model. π [PUBDEV-7748] - Target Encoding now supports feature interactions. [PUBDEV-7805] - Added CoxPH concordance to both Flow and R/Python CoxPH summaries. [PUBDEV-7820] - Added a
topbasemodel
attribute to AutoML. [PUBDEV-7831] - Added new learning curve plotting function to R/Python. [PUBDEV-7854] - Added script for estimating the memory usage of a dataset. [PUBDEV-7859] - Added fault protections to grid search allowing saving of data and parameters, model checkpointing, and auto-recovery. π [PUBDEV-7884] - Added support for Java 15. π [PUBDEV-7969] - Added CDP7.1 support. π¨ [PUBDEV-7978] - Added support for XGBoost to Print MOJO as JSON. π [PUBDEV-8021] - Added support for refreshing HDFS delegation tokens. βͺ [PUBDEV-8035] - Reverted XGBoost categorical encodings for contributions.Task
[PUBDEV-7637] -
max_hit_ratio_k
deprecated and removed. π¦ [PUBDEV-7894] - Added upper bound cap to supported Java version in H2O CRAN package requirements.π Improvement
[PUBDEV-7473] - Users now allowed to include categorical column name in beta constraints. [PUBDEV-7579] - Multinomial PDP can now be plotted for more than one target class in Flow. [PUBDEV-7736] - Sped up CoxPH concordance score by using tree instead of the direct approach. [PUBDEV-7819] - XGBoost no longer fails when specifying custom
fold_column
. [PUBDEV-7843] - XGBoost CV models now built on multiple GPUs in parallel. [PUBDEV-7968] - Missing metrics added to GLM scoring history. [PUBDEV-8017] - Added validation checks for sampling rates for XGBoost for the R/Python clients. [PUBDEV-8024] -
No longer errors when trying to use a fold column where not all folds are represented. [PUBDEV-8032] - Added themetalearner_transform
option to Stacked Ensemble. [PUBDEV-8057] - GBM main model now built in parallel to the CV models. π [PUBDEV-8060] - Removed redundant extraction weights from GBM/DRF histogram. [PUBDEV-8061] - GBM now avoids scoring the last iteration twice when early stopping is enabled. [PUBDEV-8063] - POJO predictions for XGBoost now even closer to in-H2O predictions. [PUBDEV-8064] - Double-scoring of CV models in AutoML now avoided thus speeding up AutoML. [PUBDEV-8070] - AutoML now uses fewer neurons in DL grids and has improved the metalearner for Stacked Ensemble.Technical task
[PUBDEV-7860] - Thin plate regression splines added to GAM.
π Docs
[PUBDEV-7917] - Added checkpoint description to GLM. π [PUBDEV-7976] - Added thin plate regression spline documentation to GAM algorithm page. [PUBDEV-7988] - Added missing parameters to XGBoost algorithm page. π² [PUBDEV-7992] - Added more information about log files to User Guide.