SML-Bench (Structured Machine Learning Benchmark) is a benchmark framework for machine learning from structured data. It provides datasets, which contain structured knowledge (beyond plain feature vectors) in languages such as the Web Ontology Language (OWL) or the logic programming language Prolog. For those datasets, SML-Bench defines a number of machine learning tasks, e.g. the prediction of diseases.
The ultimate goal of SML-Bench is to foster research in machine learning from structured data as well as increase the reproducibility and comparability of algorithms in that area. This is important, since a) the preparation of machine learning tasks in that area involves a significant amount of work and b) there are hardly any comparisons across languages as this requires data conversion processes.
Supported Learning Systems
SML-Bench supports several structured machine learning systems, covering ILP and the field of learning on ontologies. The supported learning systems are
- Aleph (Version 5)
- Aleph (SWI Prolog port) (Version 5)
- DL-Learner (v1.3-SNAPSHOT)
- the GILPS tools (2009-03-13), i.e
- Golem ('August 1992 version')
- Progol (5.0)