How to Represent Part-Whole Hierarchies in a Neural Network
Neural Computation · 17.02.2023 · 35 (3) · 413-452
DOI: 10.1162/neco_a_01557 ↗Kurz gefasst
Fachartikel in der Zeitschrift Neural Computation, online erschienen am 17.02.2023. Verfasst von Geoffrey Hinton. Beteiligte Einrichtung: Google Research. Die Arbeit wurde bisher 74-mal zitiert (Crossref) und nennt 64 Literaturangaben.
Abstract (Originalsprache)
This article does not describe a working system. Instead, it presents a single idea about representation that allows advances made by several different groups to be combined into an imaginary system called GLOM.1 The advances include transformers, neural fields, contrastive representation learning, distillation, and capsules. GLOM answers the question: How can a neural network with a fixed architecture parse an image into a part-whole hierarchy that has a different structure for each image? The idea is simply to use islands of identical vectors to represent the nodes in the parse tree. If GLOM can be made to work, it should significantly improve the interpretability of the representations produced by transformer-like systems when applied to vision or language.
Autor:innen und Institutionen
- Google Research.
- Vector Institute, Toronto, Ontario M5G 1M1, Canada.
- Department of Computer Science, University of Toronto, Toronto, ON M5S 2E4, Canada hinton@cs.toronto.edu.
Zitieren
APA
Hinton, G. (2023). How to Represent Part-Whole Hierarchies in a Neural Network. Neural Computation, 35(3), 413-452. https://doi.org/10.1162/neco_a_01557
Vancouver
Hinton G. How to Represent Part-Whole Hierarchies in a Neural Network. Neural Computation. 2023;35(3):413-452. doi:10.1162/neco_a_01557
BibTeX
@article{Hinton2023,
author = {Hinton, Geoffrey},
title = {How to Represent Part-Whole Hierarchies in a Neural Network},
journal = {Neural Computation},
year = {2023},
volume = {35},
number = {3},
pages = {413--452},
doi = {10.1162/neco_a_01557}
}
EUVIMED