SML-Bench

Description

SML-Bench (Structured Machine Learning Benchmark) is a benchmark 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.

A quick intro for running the base framework is given in the Getting Started document.

Mission

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 cross comparisions across languages as this requires data conversion processes.

Requirements

For Golem:

• golem binary in PATH or learningsystems/golem/Linux-x86_64
• SWI Prolog as swipl
• Python 2 as python

For Aleph:

• yap in PATH or learningsystems/aleph/Linux-x86_64
• aleph.pl in learningsystems/aleph
• Python 2 as python

For the SWI Prolog port of Aleph:

• SWI Prolog as swipl

For DL-Learner:

• Version 1.3-SNAPSHOT or higher

Supported Tools & Adding your own Tool

An overview of the currently supported tools and a brief description of how to add additional tools is given here.