mlr-org/mlr3

Built on R6 object-oriented classes and data.table for efficient computation, mlr3 provides modular building blocks for tasks, learners, and resampling strategies that can be composed and extended through a rich ecosystem of packages like mlr3tuning, mlr3pipelines, and mlr3learners. The architecture prioritizes extensibility by keeping the core lightweight while delegating specialized functionality (visualization, additional learners, autoML) to optional extension packages. Integrates with standard R workflows via checkmate for type safety and supports diverse learner backends through a unified interface.

1,060 stars. Actively maintained with 9 commits in the last 30 days.