Rotating Features for Object Discovery

TL;DR

This paper introduces Rotating Features, a scalable method for extracting object representations from neural networks, addressing the binding problem by extending complex autoencoders to higher dimensions and real-world data.

Contribution

We propose Rotating Features, a novel generalization of complex-valued features, along with a new evaluation method, enabling the use of distributed object representations in complex, real-world scenarios.

Findings

Effective extraction of objects from distributed representations

Scalability from toy data to real-world datasets

Potential to inspire new approaches to the binding problem

Abstract

The binding problem in human cognition, concerning how the brain represents and connects objects within a fixed network of neural connections, remains a subject of intense debate. Most machine learning efforts addressing this issue in an unsupervised setting have focused on slot-based methods, which may be limiting due to their discrete nature and difficulty to express uncertainty. Recently, the Complex AutoEncoder was proposed as an alternative that learns continuous and distributed object-centric representations. However, it is only applicable to simple toy data. In this paper, we present Rotating Features, a generalization of complex-valued features to higher dimensions, and a new evaluation procedure for extracting objects from distributed representations. Additionally, we show the applicability of our approach to pre-trained features. Together, these advancements enable us to scale distributed object-centric representations from simple toy to real-world data. We believe this work advances a new paradigm for addressing the binding problem in machine learning and has the potential to inspire further innovation in the field.