Learning Big Logical Rules by Joining Small Rules
Authors: Céline Hocquette, Andreas Niskanen, Rolf Morel, Matti Järvisalo, Andrew Cropper
IJCAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our experiments on many domains, including game playing and drug design, show that our approach can (i) learn rules with more than 100 literals, and (ii) drastically outperform existing approaches in terms of predictive accuracies. |
| Researcher Affiliation | Academia | 1University of Oxford 2University of Helsinki {celine.hocquette, rolf.morel, andrew.cropper}@cs.ox.ac.uk {andreas.niskanen, matti.jarvisalo}@helsinki.fi |
| Pseudocode | Yes | Algorithm 1 shows our JOINER algorithm. |
| Open Source Code | Yes | The experimental code and data are available at https://github.com/ celinehocquette/ijcai24-joiner. |
| Open Datasets | Yes | IGGP. In inductive general game playing (IGGP) [Cropper et al., 2020]... Zendo. [Bramley et al., 2018; Cropper and Hocquette, 2023]... IMDB. The international movie database (IMDB) [Mihalkova et al., 2007]... Pharmacophores. [Finn et al., 1998]... 1D-ARC. This dataset [Xu et al., 2023] contains visual reasoning tasks inspired by the abstraction and reasoning corpus [Chollet, 2019]. |
| Dataset Splits | No | The paper mentions 'positive (E+) and negative (E-) examples' and 'training examples' but does not specify the explicit percentages or counts for training, validation, and test splits for the datasets used in the experiments. |
| Hardware Specification | Yes | We use an 8-core 3.2 GHz Apple M1 and a single CPU to run the experiments. |
| Software Dependencies | Yes | We use the Max SAT solver UWr Max Sat [Piotr ow, 2020] and the SAT solver Ca Di Ca L 1.5.3 [Biere et al., 2023] (via Py SAT [Ignatiev et al., 2018]) in the join stage of JOINER. |
| Experiment Setup | Yes | We use 60s and 600s timeouts. We repeat each experiment 5 times. |