Learning Relational Causal Models with Cycles through Relational Acyclification
Authors: Ragib Ahsan, David Arbour, Elena Zheleva
AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We present experimental results to support our claim. We provide experimental results on synthetic relational models in support of our claims. We also demonstrate the effectiveness of the algorithm on a real-world dataset. Experiments In this section, we examine the effectiveness of RCD for cyclic RCMs using both synthetically generated cyclic RCMs satisfying relational acyclification criteria and a demonstration with a real-world dataset. Evaluation The goal of the evaluation is to compare the learned causal models with the true causal models. |
| Researcher Affiliation | Collaboration | Ragib Ahsan1, David Arbour2, Elena Zheleva1 1University of Illinois at Chicago, Chicago, IL 2Adobe Research 1{rahsan3, ezheleva}@uic.edu, 2arbour@adobe.com |
| Pseudocode | No | No explicit pseudocode or algorithm block was found in the paper. |
| Open Source Code | Yes | Code available at https://github.com/edgeslab/sRCD |
| Open Datasets | Yes | We generate 100 random cyclic relational causal models over randomly generated schema for each of the following combinations: entities (1 3); relationships (one less than the number of entities) with cardinalities selected uniformly at random; attributes per item drawn from Pois(λ = 1) + 1; and the number of relational dependencies (4, 6, 8, 10, 12) limited by a hop threshold of 2 and at most 3 parents per variable. Maier et al. (2013) show the output of RCD on a sample of Movie Lens dataset (www.grouplens.org) based on an approximate conditional independence test using the significance of coefficients in linear regression. |
| Dataset Splits | No | No specific details about train/validation/test splits, percentages, or cross-validation were found in the paper. |
| Hardware Specification | No | No specific hardware details (such as GPU/CPU models, memory, or cloud instance types) used for running the experiments were provided in the paper. |
| Software Dependencies | No | The paper mentions various algorithms and models (e.g., PC, FCI, RCD, Bayesian networks) but does not provide specific version numbers for any software dependencies. |
| Experiment Setup | No | The 'Experimental Setup' section describes the generation of synthetic data, including the number of entities, relationships, attributes, relational dependencies, hop threshold, and maximum parents per variable. However, it does not specify concrete hyperparameter values or training configurations for any learning process. |