Probabilistic Abduction for Visual Abstract Reasoning via Learning Rules in Vector-Symbolic Architectures
Abstract
Abstract reasoning is a cornerstone of human intelligence, and replicating it with artificial intelligence (AI) presents an ongoing challenge. This study focuses on efficiently solving Raven's progressive matrices (RPM), a visual test for assessing abstract reasoning abilities, by using distributed computation and operators provided by vector-symbolic architectures (VSA). Instead of hard-coding the rule formulations associated with RPMs, our approach can learn the VSA rule formulations (hence the name Learn-VRF) with just one pass through the training data. Yet, our approach, with compact parameters, remains transparent and interpretable. Learn-VRF yields accurate predictions on I-RAVEN's in-distribution data, and exhibits strong out-of-distribution capabilities concerning unseen attribute-rule pairs, significantly outperforming pure connectionist baselines including large language models. Our code is available at https://github.com/IBM/learn-vector-symbolic-architectures-rule-formulations.
Cite
Text
Hersche et al. "Probabilistic Abduction for Visual Abstract Reasoning via Learning Rules in Vector-Symbolic Architectures." NeurIPS 2023 Workshops: MATH-AI, 2023.Markdown
[Hersche et al. "Probabilistic Abduction for Visual Abstract Reasoning via Learning Rules in Vector-Symbolic Architectures." NeurIPS 2023 Workshops: MATH-AI, 2023.](https://mlanthology.org/neuripsw/2023/hersche2023neuripsw-probabilistic/)BibTeX
@inproceedings{hersche2023neuripsw-probabilistic,
title = {{Probabilistic Abduction for Visual Abstract Reasoning via Learning Rules in Vector-Symbolic Architectures}},
author = {Hersche, Michael and di Stefano, Francesco and Hofmann, Thomas and Sebastian, Abu and Rahimi, Abbas},
booktitle = {NeurIPS 2023 Workshops: MATH-AI},
year = {2023},
url = {https://mlanthology.org/neuripsw/2023/hersche2023neuripsw-probabilistic/}
}