OpenML is an open ecosystem for machine learning. By organizing all resources and results online, research becomes more efficient, useful and fun.
Dive into millions of results on hundreds of datasets, algorithms and workflows.
With tasks, you can share experiments on OpenML. Similar to data mining challenges, they contain data and all details that fully define an experiment, e.g., training- and test splits for cross-validation. They are downloaded from OpenML, and answered by uploading the requested results together with the algorithm or workflow used. OpenML then organizes everything online.
OpenML supports various types of tasks, sometimes with additional server-side support, but new types can be defined by users. The tasks themselves are automatically generated for new datasets, and can be searched online.
All your data is organized and linked online, so you can access everything anytime, from anywhere.
All pertinent experiment details are saved for future reference and reproducibility. Results are linked to precise implementations, versions, and parameter settings for clear analysis.
In journals, experiments are static, summarized, and scattered. Here, they are linked together, up to date, and stored in full detail.