Reorganize algorithm parameters: GBM

[exalate-issue-sync[bot]]

{noformat}Defining a GBM Model

Parameters are optional unless specified as *required*.

Algorithm-specific parameters
'''''''''''''''''''''''''''''

-  `learn_rate_annealing <algo-params/learn_rate_annealing.html>`__:  Specifies to reduce the ``learn_rate`` by this factor after every tree. So for *N* trees, GBM starts with ``learn_rate`` and ends with ``learn_rate`` :math:`\times` ``learn_rate_annealing`` :math:`^N`. For example, instead of using ``learn_rate=0.01``, you can now try ``learn_rate=0.05`` and ``learn_rate_annealing=0.99``. This method would converge much faster with almost the same accuracy. Use caution not to overfit. This value defaults to ``1``.

-  `quantile_alpha <algo-params/quantile_alpha.html>`__: (Only applicable if ``distribution="quantile"``) Specify the quantile to be used for Quantile Regression. This value has a range of 0 to 1 and defaults to ``0.5``.

-  `huber_alpha <algo-params/huber_alpha.html>`__: Specify the desired quantile for Huber/M-regression (the threshold between quadratic and linear loss). This value must be between 0 and 1 and defaults to ``0.9``.

-  `pred_noise_bandwidth <algo-params/pred_noise_bandwidth.html>`__: The bandwidth (sigma) of Gaussian multiplicative noise ~N(1,sigma) for tree node predictions. If this parameter is specified with a value greater than zero, then every leaf node prediction is randomly scaled by a number drawn from a Normal distribution centered around 1 with a bandwidth given by this parameter. The default is ``0`` (disabled). 

- `upload_custom_distribution <algo-params/upload_custom_distribution.html>`__: Upload a custom distribution into a running H2O cluster.

-  `custom_distribution_func <algo-params/custom_distribution_func.html>`__: Specify a custom distribution function. A demo for the custom distribution function is `available here <https://github.com/h2oai/h2o-3/blob/master/h2o-py/demos/custom_loss_function_demo.ipynb>`__.

-  `in_training_checkpoints_dir <algo-params/in_training_checkpoints_dir.html>`__: Create checkpoints in the defined directory while the training process is still running. In case of a cluster shutdown, this checkpoint can be used to restart training.

-  `in_training_checkpoints_tree_interval <algo-params/in_training_checkpoints_tree_interval.html>`__: Checkpoint the model after every so many trees. Parameter is used only when ``in_training_checkpoints_dir`` is defined. The default value is ``1`` and makes the checkpoint after each trained tree.

Tree-based algorithm parameters
'''''''''''''''''''''''''''''''

-  `score_tree_interval <algo-params/score_tree_interval.html>`__: Score the model after every so many trees. This is set to ``0`` (disabled) by default.

-  `ntrees <algo-params/ntrees.html>`__: Specify the number of trees to build (defaults to ``50``).

-  `max_depth <algo-params/max_depth.html>`__: Specify the maximum tree depth. Higher values will make the model more complex and can lead to overfitting. Setting this value to ``0`` specifies no limit. This value defaults to ``5``.

-  `min_rows <algo-params/min_rows.html>`__: Specify the minimum number of observations for a leaf (``nodesize`` in R). This value defaults to ``10``.

-  `nbins <algo-params/nbins.html>`__: (Numerical/real/int only) Specify the number of bins for the histogram to build, then split at the best point (defaults to ``20``).

-  `nbins_top_level <algo-params/nbins_top_level.html>`__: (For numerical/real/int columns only) Specify the minimum number of bins at the root level to use to build the histogram. This number will then be decreased by a factor of two per level. This value defaults to ``1024``.

-  `nbins_cats <algo-params/nbins_cats.html>`__: (Categorical/enums only) Specify the maximum number of bins for the histogram to build, then split at the best point. Higher values can lead to more overfitting. The levels are ordered alphabetically; if there are more levels than bins, adjacent levels share bins. This value has a more significant impact on model fitness than ``nbins``. Larger values may increase runtime, especially for deep trees and large clusters, so tuning may be required to find the optimal value for your configuration. This value defaults to ``1024``.

-  `col_sample_rate <algo-params/col_sample_rate.html>`__: Specify the column sampling rate (y-axis). (Note that this method is sampling without replacement.) The range is 0.0 to 1.0, and this value defaults to ``1``. Higher values may improve training accuracy. Test accuracy improves when either columns or rows are sampled. For details, refer to "Stochastic Gradient Boosting" (`Friedman, 1999 <https://statweb.stanford.edu/~jhf/ftp/stobst.pdf>`__).

-  `max_abs_leafnode_pred <algo-params/max_abs_leafnode_pred.html>`__: When building a GBM classification model, this option reduces overfitting by limiting the maximum absolute value of a leaf node prediction. This option defaults to ``1.797693135e+308``.

-  `learn_rate <algo-params/learn_rate.html>`__: Specify the learning rate. The range is 0.0 to 1.0, and the default value is ``0.1``.

-  `build_tree_one_node <algo-params/build_tree_one_node.html>`__: Specify whether to run on a single node. This is suitable for small datasets as there is no network overhead but fewer CPUs are used. This option defaults to ``False`` (disabled).

-  `sample_rate <algo-params/sample_rate.html>`__: Specify the row sampling rate (x-axis). (Note that this method samples without replacement.) The range is ``0.0`` to ``1.0``, and this value defaults to ``1``. Higher values may improve training accuracy. Test accuracy improves when either columns or rows are sampled. For details, refer to "Stochastic Gradient Boosting" (`Friedman, 1999 <https://statweb.stanford.edu/~jhf/ftp/stobst.pdf>`__).

-  `sample_rate_per_class <algo-params/sample_rate_per_class.html>`__: When building models from imbalanced datasets, this option specifies that each tree in the ensemble should sample from the full training dataset using a per-class-specific sampling rate rather than a global sample factor (as with ``sample_rate``). The range for this option is 0.0 to 1.0. Note that this method samples without replacement.

-  `col_sample_rate_change_per_level <algo-params/col_sample_rate_change_per_level.html>`__: This option specifies to change the column sampling rate as a function of the depth in the tree. This can be a value > 0.0 and :math:`\leq` 2.0 and defaults to ``1``. (Note that this method is sampling without replacement.) For example:

   - **level 1**: :math:`\text{col_sample_rate}`
   - **level 2**: :math:`\text{col_sample_rate} \times \text{factor}` 
   - **level 3**: :math:`\text{col_sample_rate} \times \text{factor}^2`
   - **level 4**: :math:`\text{col_sample_rate} \times \text{factor}^3`
   - etc.

-  `col_sample_rate_per_tree <algo-params/col_sample_rate_per_tree.html>`__: Specify the column sample rate per tree. This can be a value from 0.0 to 1.0 and defaults to ``1``. Note that it is multiplicative with ``col_sample_rate``, so setting both parameters to ``0.8``, for example, results in 64% of columns being considered at any given node to split. Note that this method is sampling without replacement.

-  `min_split_improvement <algo-params/min_split_improvement.html>`__: The value of this option specifies the minimum relative improvement in squared error reduction in order for a split to happen. When properly tuned, this option can help reduce overfitting. Optimal values would be in the 1e-10 to 1e-3 range, and this value defaults to ``1e-05``.

-  `histogram_type <algo-params/histogram_type.html>`__: By default (``AUTO``), GBM bins from min to max in steps of :math:`\frac{(\text{max-min})}{N}`. ``Random`` split points or quantile-based split points can be selected as well. ``RoundRobin`` can be specified to cycle through all histogram types (one per tree). Use this option to specify the type of histogram to use for finding optimal split points. One of:

   - ``AUTO``
   - ``UniformAdaptive``
   - ``Random``
   - ``QuantilesGlobal``
   - ``RoundRobin``

-  `calibrate_model <algo-params/calibrate_model.html>`__: Use Platt scaling to calculate calibrated class probabilities. This option defaults to ``False`` (disabled).

-  `calibration_frame <algo-params/calibration_frame.html>`__: Specifies the frame to be used for Platt scaling.

- **calibration_method**: Calibration method to use. Must be one of: ``"auto"`` (default), ``"platt_scaling"``, or ``"isotonic_regression"``.

- `upload_custom_metric <algo-params/upload_custom_metric.html>`__: Upload a custom metric into a running H2O cluster.

-  `custom_metric_func <algo-params/custom_metric_func.html>`__: Specify a custom evaluation function.

-  `check_constant_response <algo-params/check_constant_response.html>`__: Check if the response column is a constant value. If enabled (default), then an exception is thrown if the response column is a constant value. If disabled, then the model will train regardless of the response column being a constant value or not. This option defaults to ``True`` (enabled).

- `interaction_constraints <algo-params/interaction_constraints.html>`__: A set of allowed column interactions. This option defaults to ``None`` which means all column features are included. This option can only be used when ``categorical_encoding=AUTO``.

Common parameters
'''''''''''''''''

-  `training_frame <algo-params/training_frame.html>`__: *Required* Specify the dataset used to build the model. **NOTE**: In Flow, if you click the **Build a model** button from the ``Parse`` cell, the training frame is entered automatically.

-  `y <algo-params/y.html>`__: *Required* Specify the column to use as the dependent variable. The data can be numeric or categorical.

-  `x <algo-params/x.html>`__: Specify a vector containing the names or indices of the predictor variables to use when building the model. If ``x`` is missing, then all columns except ``y`` are used.

-  `validation_frame <algo-params/validation_frame.html>`__: Specify the dataset used to evaluate the accuracy of the model.

-  `model_id <algo-params/model_id.html>`__: Specify a custom name for the model to use as a reference. By default, H2O automatically generates a destination key.

-  `nfolds <algo-params/nfolds.html>`__: Specify the number of folds for cross-validation. The value can be ``0`` (default) to disable or :math:`\geq` ``2``.

-  `keep_cross_validation_models <algo-params/keep_cross_validation_models.html>`__: Specify whether to keep the cross-validated models. Keeping cross-validation models may consume significantly more memory in the H2O cluster. This option defaults to ``True`` (enabled).

-  `keep_cross_validation_predictions <algo-params/keep_cross_validation_predictions.html>`__: Enable this option to keep the cross-validation predictions. This option defaults to ``False`` (disabled).

-  `keep_cross_validation_fold_assignment <algo-params/keep_cross_validation_fold_assignment.html>`__: Enable this option to preserve the cross-validation fold assignment. This option defaults to ``False`` (disabled). 

-  `score_each_iteration <algo-params/score_each_iteration.html>`__: Specify whether to score during each iteration of the model training. This value is set to ``False`` (disabled) by default.

-  `fold_assignment <algo-params/fold_assignment.html>`__: (Applicable only if a value for ``nfolds`` is specified and ``fold_column`` is not specified) Specify the cross-validation fold assignment scheme. One of:

   - ``AUTO`` (default; uses ``Random``)
   - ``Random``
   - ``Modulo`` (read more `here <https://en.wikipedia.org/wiki/Modulo_operation>`__)
   - ``Stratified`` (which will stratify the folds based on the response variable for classification problems)

-  `fold_column <algo-params/fold_column.html>`__: Specify the column that contains the cross-validation fold index assignment per observation.

-  `ignored_columns <algo-params/ignored_columns.html>`__: (Python and Flow only) Specify the column or columns to be excluded from the model. In Flow, click the checkbox next to a column name to add it to the list of columns excluded from the model. To add all columns, click the **All** button. To remove a column from the list of ignored columns, click the X next to the column name. To remove all columns from the list of ignored columns, click the **None** button. To search for a specific column, type the column name in the **Search** field above the column list. To only show columns with a specific percentage of missing values, specify the percentage in the **Only show columns with more than 0% missing values** field. To change the selections for the hidden columns, use the **Select Visible** or **Deselect Visible** buttons.

-  `ignore_const_cols <algo-params/ignore_const_cols.html>`__: Specify whether to ignore constant training columns, since no information can be gained from them. This option defaults to ``True`` (enabled).

-  `offset_column <algo-params/offset_column.html>`__: (Not applicable if the ``distribution`` is ``multinomial``) Specify a column to use as the offset.

    **Note**: Offsets are per-row "bias values" that are used during model training. For Gaussian distributions, they can be seen as simple corrections to the response (``y``) column. Instead of learning to predict the response (y-row), the model learns to predict the (row) offset of the response column. For other distributions, the offset corrections are applied in the linearized space before applying the inverse link function to get the actual response values. 

-  `weights_column <algo-params/weights_column.html>`__: Specify a column to use for the observation weights, which are used for bias correction. The specified ``weights_column`` must be included in the specified ``training_frame``. 

    *Python only*: To use a weights column when passing an H2OFrame to ``x`` instead of a list of column names, the specified ``training_frame`` must contain the specified ``weights_column``. 

    **Note**: Weights are per-row observation weights and do not increase the size of the data frame. This is typically the number of times a row is repeated, but non-integer values are supported as well. During training, rows with higher weights matter more, due to the larger loss function pre-factor.

-  `balance_classes <algo-params/balance_classes.html>`__: Specify whether to oversample the minority classes to balance the class distribution. This option defaults to ``False`` (disabled), and can increase the data frame size. This option is only applicable for classification. Majority classes can be undersampled to satisfy the ``max_after_balance_size`` parameter.

-  `class_sampling_factors <algo-params/class_sampling_factors.html>`__: Specify the per-class (in lexicographical order) over/under-sampling ratios. By default, these ratios are automatically computed during training to obtain the class balance. Note that this requires ``balance_classes=true``.

-  `max_after_balance_size <algo-params/max_after_balance_size.html>`__: Specify the maximum relative size of the training data after balancing class counts (``balance_classes`` must be enabled). The value can be less than 1.0 and defaults to ``5.0``.

-  `stopping_metric <algo-params/stopping_metric.html>`__: Specify the metric to use for early stopping. The available options are:

    - ``AUTO``: This defaults to ``logloss`` for classification, ``deviance`` for regression, and ``anomaly_score`` for Isolation Forest. 
    - ``anomaly_score`` (Isolation Forest only)
    - ``deviance``
    - ``logloss``
    - ``MSE``
    - ``RMSE``
    - ``MAE``
    - ``RMSLE``
    - ``AUC`` (area under the ROC curve)
    - ``AUCPR`` (area under the Precision-Recall curve)
    - ``lift_top_group`` 
    - ``misclassification``
    - ``mean_per_class_error``
    - ``custom`` (Python client only)
    - ``custom_increasing`` (Python client only)

  Note that ``custom`` and ``custom_increasing`` can only be used in GBM and DRF with the Python client.   

-  `stopping_rounds <algo-params/stopping_rounds.html>`__: Stops training when the option selected for ``stopping_metric`` doesn't improve for the specified number of training rounds, based on a simple moving average. This option defaults ``0`` (no early stopping). The metric is computed on the validation data (if provided); otherwise, training data is used. **Note**: If cross-validation is enabled:

    - All cross-validation models stop training when the validation metric doesn't improve.
    - The main model runs for the mean number of epochs.
    - N+1 models may be off by the number specified for ``stopping_rounds`` from the best model, but the cross-validation metric estimates the performance of the main model for the resulting number of epochs (which may be fewer than the specified number of epochs).

-  `stopping_tolerance <algo-params/stopping_tolerance.html>`__: Specify the relative tolerance for the metric-based stopping to stop training if the improvement is less than this value. This value defaults to ``0.001``.

-  `max_runtime_secs <algo-params/max_runtime_secs.html>`__: Maximum allowed runtime in seconds for model training.  This defaults to ``0`` (unlimited).

-  `seed <algo-params/seed.html>`__: Specify the random number generator (RNG) seed for algorithm components dependent on randomization. The seed is consistent for each H2O instance so that you can create models with the same starting conditions in alternative configurations. This value defaults to ``-1`` (time-based random number).

-  `distribution <algo-params/distribution.html>`__: Specify the distribution (i.e., the loss function). The options are ``AUTO`` (default), ``bernoulli``, ``multinomial``, ``gaussian``, ``poisson``, ``gamma``, ``laplace``, ``quantile``, ``huber``, or ``tweedie``.

  - If the distribution is ``bernoulli``, the the response column must be 2-class categorical.
  - If the distribution is ``quasibinomial``, the response column must be numeric and binary. 
  - If the distribution is ``multinomial``, the response column must be categorical.
  - If the distribution is ``poisson``, the response column must be numeric.
  - If the distribution is ``laplace``, the response column must be numeric.
  - If the distribution is ``tweedie``, the response column must be numeric.
  - If the distribution is ``gaussian``, the response column must be numeric.
  - If the distribution is ``huber``, the response column must be numeric.
  - If the distribution is ``gamma``, the response column must be numeric.
  - If the distribution is ``quantile``, the response column must be numeric.

-  `tweedie_power <algo-params/tweedie_power.html>`__: (Only applicable if ``distribution="tweedie"``) Specify the Tweedie power. You can tune over this option with values > 1.0 and < 2.0. This value defaults to ``1.5``. For more information, refer to `Tweedie distribution <https://en.wikipedia.org/wiki/Tweedie_distribution>`__.

-  `checkpoint <algo-params/checkpoint.html>`__: Enter a model key associated with a previously trained model. Use this option to build a new model as a continuation of a previously generated model.

- `categorical_encoding <algo-params/categorical_encoding.html>`__: Specify one of the following encoding schemes for handling categorical features:

  - ``auto`` or ``AUTO``: Allow the algorithm to decide (default). In GBM, the algorithm will automatically perform ``enum`` encoding.
  - ``enum`` or ``Enum``: 1 column per categorical feature.
  - ``enum_limited`` or ``EnumLimited``: Automatically reduce categorical levels to the most prevalent ones during training and only keep the **T** (10) most frequent levels.
  - ``one_hot_explicit`` or ``OneHotExplicit``: N+1 new columns for categorical features with N levels.
  - ``binary``: No more than 32 columns per categorical feature.
  - ``eigen`` or ``Eigen``: *k* columns per categorical feature, keeping projections of one-hot-encoded matrix onto *k*-dim eigen space only.
  - ``label_encoder`` or ``LabelEncoder``:  Convert every enum into the integer of its index (for example, level 0 -> 0, level 1 -> 1, etc.).
  - ``sort_by_response`` or ``SortByResponse``: Reorders the levels by the mean response (for example, the level with lowest response -> 0, the level with second-lowest response -> 1, etc.). This is useful in GBM/DRF, for example, when you have more levels than ``nbins_cats``, and where the top level splits now have a chance at separating the data with a split. Note that this requires a specified response column.

-  `export_checkpoints_dir <algo-params/export_checkpoints_dir.html>`__: Specify a directory to which generated models will automatically be exported.

-  `monotone_constraints <algo-params/monotone_constraints.html>`__: A mapping representing monotonic constraints. Use ``+1`` to enforce an increasing constraint and ``-1`` to specify a decreasing constraint. Note that constraints can only be defined for numerical columns. Also note that this option can only be used when the distribution is ``gaussian``, ``bernoulli``, ``tweedie`` or ``quantile``. A Python demo is available `here <https://github.com/h2oai/h2o-3/tree/master/h2o-py/demos/H2O_tutorial_gbm_monotonicity.ipynb>`__.

- `gainslift_bins <algo-params/gainslift_bins.html>`__: The number of bins for a Gains/Lift table. The default value is ``-1`` and makes the binning automatic. To disable this feature, set to ``0``.

- `auc_type <algo-params/auc_type.html>`__: Set the default multinomial AUC type. Must be one of:

    - ``"AUTO"`` (default)
    - ``"NONE"``
    - ``"MACRO_OVR"``
    - ``"WEIGHTED_OVR"``
    - ``"MACRO_OVO"``
    - ``"WEIGHTED_OVO"``

-  **verbose**: Print scoring history to the console. For GBM, metrics are per tree. This option defaults to ``False`` (disabled).
 {noformat}

[h2o-ops]

JIRA Issue Details

Jira Issue: PUBDEV-9059 Assignee: hannah.tillman Reporter: hannah.tillman State: Resolved Fix Version: 3.40.0.4 Attachments: N/A Development PRs: Available

[h2o-ops]

Linked PRs from JIRA

https://github.com/h2oai/h2o-3/pull/6691