Recurrent Complex-Weighted Autoencoders for Unsupervised Object Discovery

TL;DR

This paper introduces SynCx, a recurrent complex-weighted autoencoder that improves unsupervised object discovery by iteratively enforcing phase-based feature binding, outperforming existing models and reducing grouping errors.

Contribution

The paper presents SynCx, a novel recurrent autoencoder with complex weights that performs iterative constraint satisfaction for unsupervised object discovery, eliminating the need for additional binding mechanisms.

Findings

SynCx outperforms current models in unsupervised object discovery tasks.

SynCx avoids systematic grouping errors like confusing similarly colored objects.

SynCx is strongly competitive with state-of-the-art models.

Abstract

Current state-of-the-art synchrony-based models encode object bindings with complex-valued activations and compute with real-valued weights in feedforward architectures. We argue for the computational advantages of a recurrent architecture with complex-valued weights. We propose a fully convolutional autoencoder, SynCx, that performs iterative constraint satisfaction: at each iteration, a hidden layer bottleneck encodes statistically regular configurations of features in particular phase relationships; over iterations, local constraints propagate and the model converges to a globally consistent configuration of phase assignments. Binding is achieved simply by the matrix-vector product operation between complex-valued weights and activations, without the need for additional mechanisms that have been incorporated into current synchrony-based models. SynCx outperforms or is strongly competitive with current models for unsupervised object discovery. SynCx also avoids certain systematic grouping errors of current models, such as the inability to separate similarly colored objects without additional supervision.