Learn R Programming

lightgbm (version 4.5.0)

lgb_shared_params: Shared parameter docs

Description

Parameter docs shared by lgb.train, lgb.cv, and lightgbm

Arguments

callbacks

List of callback functions that are applied at each iteration.

data

a lgb.Dataset object, used for training. Some functions, such as lgb.cv, may allow you to pass other types of data like matrix and then separately supply label as a keyword argument.

early_stopping_rounds

int. Activates early stopping. When this parameter is non-null, training will stop if the evaluation of any metric on any validation set fails to improve for early_stopping_rounds consecutive boosting rounds. If training stops early, the returned model will have attribute best_iter set to the iteration number of the best iteration.

eval

evaluation function(s). This can be a character vector, function, or list with a mixture of strings and functions.

  • a. character vector: If you provide a character vector to this argument, it should contain strings with valid evaluation metrics. See The "metric" section of the documentation for a list of valid metrics.

  • b. function: You can provide a custom evaluation function. This should accept the keyword arguments preds and dtrain and should return a named list with three elements:

    • name: A string with the name of the metric, used for printing and storing results.

    • value: A single number indicating the value of the metric for the given predictions and true values

    • higher_better: A boolean indicating whether higher values indicate a better fit. For example, this would be FALSE for metrics like MAE or RMSE.

  • c. list: If a list is given, it should only contain character vectors and functions. These should follow the requirements from the descriptions above.

eval_freq

evaluation output frequency, only effective when verbose > 0 and valids has been provided

init_model

path of model file or lgb.Booster object, will continue training from this model

nrounds

number of training rounds

obj

objective function, can be character or custom objective function. Examples include regression, regression_l1, huber, binary, lambdarank, multiclass, multiclass

params

a list of parameters. See the "Parameters" section of the documentation for a list of parameters and valid values.

verbose

verbosity for output, if <= 0 and valids has been provided, also will disable the printing of evaluation during training

serializable

whether to make the resulting objects serializable through functions such as save or saveRDS (see section "Model serialization").

Early Stopping

"early stopping" refers to stopping the training process if the model's performance on a given validation set does not improve for several consecutive iterations.

If multiple arguments are given to eval, their order will be preserved. If you enable early stopping by setting early_stopping_rounds in params, by default all metrics will be considered for early stopping.

If you want to only consider the first metric for early stopping, pass first_metric_only = TRUE in params. Note that if you also specify metric in params, that metric will be considered the "first" one. If you omit metric, a default metric will be used based on your choice for the parameter obj (keyword argument) or objective (passed into params).

Model serialization

LightGBM model objects can be serialized and de-serialized through functions such as save or saveRDS, but similarly to libraries such as 'xgboost', serialization works a bit differently from typical R objects. In order to make models serializable in R, a copy of the underlying C++ object as serialized raw bytes is produced and stored in the R model object, and when this R object is de-serialized, the underlying C++ model object gets reconstructed from these raw bytes, but will only do so once some function that uses it is called, such as predict. In order to forcibly reconstruct the C++ object after deserialization (e.g. after calling readRDS or similar), one can use the function lgb.restore_handle (for example, if one makes predictions in parallel or in forked processes, it will be faster to restore the handle beforehand).

Producing and keeping these raw bytes however uses extra memory, and if they are not required, it is possible to avoid producing them by passing `serializable=FALSE`. In such cases, these raw bytes can be added to the model on demand through function lgb.make_serializable.

New in version 4.0.0