The Scattering Compositional Learner: Discovering Objects, Attributes, Relationships in Analogical Reasoning

Yuhuai Wu, Honghua Dong, Roger Baker Grosse, Jimmy Ba

Keywords: Raven's Progressive Matrices, visual analogical reasoning.

Abstract: In this work, we focus on an analogical reasoning task that contains rich compositional structures, Raven's Progressive Matrices (RPM). To discover compositional structures of the data, we propose the Scattering Compositional Learner (SCL), an architecture that composes neural networks in a sequence. Our SCL achieves state-of-the-art performance on two RPM datasets, with a 48.7% relative improvement on Balanced-RAVEN and 26.4% on PGM over the previous state-of-the-art. We additionally show that our model discovers compositional representations of objects' attributes (e.g., shape color, size), and their relationships (e.g., progression, union). We also find that the compositional representation makes the SCL significantly more robust to test-time domain shifts and greatly improves zero-shot generalization to previously unseen analogies.

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

One-sentence Summary: We designed a neural network architecture for analogical reasoning for Raven's Progressive Matrices, achieving SOTA on this task.

Community Implementations: [ 3 code implementations](https://www.catalyzex.com/paper/arxiv:2007.04212/code)

Reviewed Version (pdf): https://openreview.net/references/pdf?id=reLacWb2nm