Optimised Storage for Datalog Reasoning
Authors: Xinyue Zhang, Pan Hu, Yavor Nenov, Ian Horrocks
AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our experimental evaluation shows that our approach significantly improves memory consumption, sometimes by orders of magnitude, while remaining competitive in terms of query answering time. |
| Researcher Affiliation | Collaboration | Xinyue Zhang1, Pan Hu2, Yavor Nenov3, Ian Horrocks1 1Department of Computer Science, University of Oxford, Oxford, UK 2School of Electrical Information and Electronic Engineering, Shanghai Jiao Tong University, China 3Oxford Semantic Techonologies, Oxford, UK |
| Pseudocode | Yes | Algorithm 1: Seminaive(Π, I, E, E+) ... Algorithm 2: Multi-Scheme Reasoning(Π, E, E+) ... Algorithm 3: Functions of Union Table |
| Open Source Code | Yes | The datasets and test systems are available online2. 2https://xinyuezhang.xyz/TCReasoning/ |
| Open Datasets | Yes | Benchmarks: We tested our algorithms on DAG-R (Hu, Motik, and Horrocks 2022), DBpedia (Lehmann et al. 2015), and Relations (Smith et al. 2007). |
| Dataset Splits | No | The paper discusses using subsets of datasets (e.g., 'randomly choosing subsets from DBpedia') and mentions inserting facts in 'four rounds' for evaluation, but does not explicitly provide details about train, validation, or test dataset splits or percentages. |
| Hardware Specification | Yes | All of our experiments are conducted on a Dell Power Edge R730 server with 512GB RAM and 2 Intel Xeon E5-2640 2.60GHz processors, running Fedora 33, kernel version 5.10.8. |
| Software Dependencies | No | The paper mentions the operating system and kernel version ('Fedora 33, kernel version 5.10.8'). However, it does not provide specific version numbers for other ancillary software dependencies, such as programming languages, libraries, or frameworks (e.g., Python, PyTorch, specific Datalog engine versions used/modified beyond naming them). |
| Experiment Setup | No | The paper describes aspects of the test setup such as how facts were inserted ('inserted the facts in four rounds') and how scalability was evaluated ('randomly choosing subsets from DBpedia'). However, it does not provide specific experimental configuration details such as hyperparameters, learning rates, batch sizes, number of epochs, or specific configurable parameters for the algorithms or models. |