On Learning Causal Models from Relational Data

Authors: Sanghack Lee, Vasant Honavar

AAAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments confirm that RCDℓis substantially more efficient than RCD with respect to its space and time requirements (see Figure 4). RCDℓtakes 70 seconds on average learning an RCM given h = 4 while RCD takes 50 minutes.
Researcher Affiliation Academia Sanghack Lee and Vasant Honavar Artificial Intelligence Research Laboratory College of Information Sciences and Technology The Pennsylvania State University University Park, PA 16802 {sxl439, vhonavar}@ist.psu.edu
Pseudocode Yes Algorithm 1 RCDℓ: RCD-Light
Open Source Code No The paper does not provide an explicit statement about releasing source code or a link to a code repository for the methodology described.
Open Datasets No The paper states, 'We generated schemas with 3 entity and 3 binary relationship classes with 2 and 1 attribute classes per entity and relationship class, respectively, with random cardinality. Given the schema, we generated an RCM with 10 dependencies of length up to h and maximum degree of 3.' This indicates synthetic data generation, but no public access information or citation for a dataset is provided.
Dataset Splits No The paper describes the generation of synthetic models but does not specify any training, validation, or test dataset splits.
Hardware Specification No The paper does not provide any specific details about the hardware used to run the experiments.
Software Dependencies No The paper mentions 'RCDℓ (built on RCD codebase)' but does not specify any software names with version numbers.
Experiment Setup Yes We generated schemas with 3 entity and 3 binary relationship classes with 2 and 1 attribute classes per entity and relationship class, respectively, with random cardinality. Given the schema, we generated an RCM with 10 dependencies of length up to h and maximum degree of 3. We followed settings in Maier et al. (2013): (i) RCD uses AGGs whose hop length is limited to 2h for practical reasons; and (ii) AGGs with 2h is adopted as a CI oracle.