R: mlr3oml: Connector Between 'mlr3' and 'OpenML'

mlr3oml-package {mlr3oml}

R Documentation

mlr3oml: Connector Between 'mlr3' and 'OpenML'

Description

Provides an interface to 'OpenML.org' to list and download machine learning data, tasks and experiments. The 'OpenML' objects can be automatically converted to 'mlr3' objects. For a more sophisticated interface with more upload options, see the 'OpenML' package.

Documentation

Start by reading the Large-Scale Benchmarking chapter from the mlr3book.

mlr3 Integration

This package adds the mlr3::Task "oml" and the mlr3::Resampling "oml" to mlr3::mlr_tasks and mlr3::mlr_resamplings, respectively. For the former you may pass either a data_id or a task_id, the latter requires a task_id. Furthermore it allows to convert the OpenML objects to mlr3 objects using the usual S3 generics such as mlr3::as_task, mlr3::as_learner, mlr3::as_resampling, mlr3::as_resample_result, mlr3::as_benchmark_result or mlr3::as_data_backend. This allows for a frictionless integration of OpenML and mlr3.

Options

mlr3oml.cache: Enables or disables caching globally. If set to FALSE, caching is disabled. If set to TRUE, cache directory as reported by R_user_dir() is used. Alternatively, you can specify a path on the local file system here. Default is FALSE.
mlr3oml.api_key: API key to use. All operations supported by this package work without an API key, but you might get rate limited without an API key. If not set, defaults to the value of the environment variable OPENMLAPIKEY.
mlr3oml.arff_parser: ARFF parser to use, defaults to the internal one relies on data.table::fread(). Can also be set to "RWeka" for the parser in RWeka.
mlr3oml.parquet: Enables or disables parquet as the default file format. If set to TRUE, the parquet version of datasets will be used by default. If set to FALSE, the arff version of datasets will be used by default. Note that the OpenML sever is still transitioning from arff to parquet and some features will work better with arff. Default is FALSE.
mlr3oml.retries: An integer defining number of retries when downloading data from OpenML. If it is NULL, the number of retries is set to 3.

Relevant for developers

mlr3oml.test_server: The default value for whether to use the OpenML test server. Default is FALSE.
mlr3oml.test_api_key: API key to use for the test server. If not set, defaults to the value of the environment variable TESTOPENMLAPIKEY.

Logging

The lgr package is used for logging. To change the threshold, use lgr::get_logger("mlr3oml")$set_threshold().

Author(s)

Maintainer: Sebastian Fischer sebf.fischer@gmail.com (ORCID)

Authors:

Michel Lang michellang@gmail.com (ORCID)