This module contains experiments with RDFRules and AnyBURL.
Clone the RDFRules repository and run following SBT commands:
> project experimentsKgc
The main class RdfRulesKgc requires some parameters to determine which kind of experiment should be launched:
> runMain com.github.propi.rdfrules.experiments.RdfRulesKgc parameters...
| Parameter name | Description | Default |
|---|---|---|
| cores | Number of threads. | available cores |
| dataset | Input dataset (wn18rr, fb15k-237, yago3-10) | wn18rr |
| output | Output file with benchmark results. | experiments/data/results.txt |
| rlen | Max rule length | 3 |
| revalidate | Revalidate all intermediate files | false |
| runconfidences | Run KGC task with different confidence measures (CWA, PCA, QPCA) | |
| runmodes | Run KGC with modes (most frequent items added into each prediction task) | |
| runconstants | Run KGC without constants | |
| runanytime | Run KGC with anytime approach for rule refinement | |
| runscorers | Run KGC with various scorers (NoisyOr, NonRedundantNoisyOr, NonRedundantMaximum) | |
| runpruning | Run KGC with data coverage pruning | |
| runanyburl | Run AnyBURL algorithm for KGC. Input thresholds are same as for RDFRules. Max mining time is set to 1000s |
Default mining parameters are: minSupport = 5, minHeadSize = 1, minConfidence = 0.1
We performed some experiments in order to solve a knowledge graph completion (KGC) problem on the CESNET Metacentrum computing cluster with RDFRules 1.7.2 and AnyBURL 22 with following input parameters:
# task1 - run and evaluate the KGC task for wn18rr dataset computed by RDFRules with different confidence types.
> runMain com.github.propi.rdfrules.experiments.RdfRulesKgc -cores 8 -runconfidences
# task2 - run and evaluate the KGC task for wn18rr dataset computed by RDFRules with the data coverage pruning strategy.
> runMain com.github.propi.rdfrules.experiments.RdfRulesKgc -cores 8 -runpruning
# task3 - run and evaluate the KGC task for wn18rr dataset computed by AnyBURL.
> runMain com.github.propi.rdfrules.experiments.RdfRulesKgc -cores 8 -runanyburl
The parameters of the used machine for experiments are:
- CPU: 4x 14-core Intel Xeon E7-4830 v4 (2GHz),
- RAM: 512 GB,
- OS: Debian 9.