Recurrent Complex-Weighted Autoencoders for Unsupervised Object Discovery

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Gopalakrishnan, Anand, Stanić, Aleksandar, Schmidhuber, Jürgen, Mozer, Michael Curtis
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866909367267229696
author Gopalakrishnan, Anand
Stanić, Aleksandar
Schmidhuber, Jürgen
Mozer, Michael Curtis
author_facet Gopalakrishnan, Anand
Stanić, Aleksandar
Schmidhuber, Jürgen
Mozer, Michael Curtis
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2405_17283
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Recurrent Complex-Weighted Autoencoders for Unsupervised Object Discovery
Gopalakrishnan, Anand
Stanić, Aleksandar
Schmidhuber, Jürgen
Mozer, Michael Curtis
Machine Learning
Neural and Evolutionary Computing
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.
title Recurrent Complex-Weighted Autoencoders for Unsupervised Object Discovery
topic Machine Learning
Neural and Evolutionary Computing
url https://arxiv.org/abs/2405.17283