Year: 2023
Type: preprint
Abstract: Learning representations that generalize to novel compositions of known concepts is crucial for bridging the gap between human and machine perception. One prominent effort is learning object-centric representations, which are widely conjectured to enable compositional generalization. Yet, it remains unclear when this conjecture will be true, as a principled theoretical or empirical understanding of compositional generalization is lacking. In this work, we investigate when compositional generalization is guaranteed for object-centric representations through the lens of identifiability theory. We show that autoencoders that satisfy structural assumptions on the decoder and enforce encoder-decoder consistency will learn object-centric representations that provably generalize compositionally. We validate our theoretical result and highlight the practical relevance of our assumptions through experiments on synthetic image data.
Source: arXiv (Cornell University)
Authors: Wiedemer, Thaddäus, Brady, Jack, Panfilov, Alexander, Juhos, Attila, Bethge, Matthias, Brendel, Wieland
Institutions: Tübingen AI Center, University of Tübingen
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
Funders: International Max Planck Research School for Environmental, Cellular and Molecular Microbiology, Good Ventures Foundation, Deutsche Forschungsgemeinschaft, Bundesministerium für Bildung und Forschung, International Max Planck Research School for Advanced Methods in Process and Systems Engineering
OpenAlex ID: W4387559920