Robust Stochastic Graph Generator for Counterfactual Explanations
Keywords: explainable AI, generative AI, algorithmic recourse, graph neural networks, deep learning
Abstract: Counterfactual Explanation (CE) techniques are used to provide insights to AI system users. While well-researched in domains like medical imaging and autonomous vehicles, Graph Counterfactual Explanation (GCE) methods, especially generative ones, have been less explored. GCEs generate a new graph akin to the original one, having a different outcome grounded on the underlying predictive model. Generative approaches, despite their success in domains like artistic styles and natural language modeling, have received limited attention in GCE. We introduce RSGG-CE, a novel Robust Stochastic Graph Generator for Counterfactual Explanations capable of generating counterfactual examples from the learned latent space with a partially ordered generation sequence. Our study quantitatively and qualitatively compares RSGG-CE’s performance to state-of-the-art (SoA) generative explainers, demonstrating its superior ability to generate plausible counterfactual candidates.
Primary Area: general machine learning (i.e., none of the above)
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 5764
Loading