Higgins, I;
Sonnerat, N;
Matthey, L;
Pal, A;
Burgess, CP;
Bošnjak, M;
Shanahan, M;
... Lerchner, A; + view all
(2018)
SCAN: Learning Hierarchical Compositional Visual Concepts.
In: Bengio, Y and LeCun, Y, (eds.)
Proceedings of the Sixth International Conference on Learning Representations (ICLR 2018).
International Conference on Learning Representations (ICLR): Vancouver, Canada.
Preview |
Text
Bosjnak_scan_learning_hierarchical_compositional_visual_concepts.pdf - Published Version Download (22MB) | Preview |
Abstract
The seemingly infinite diversity of the natural world arises from a relatively small set of coherent rules, such as the laws of physics or chemistry. We conjecture that these rules give rise to regularities that can be discovered through primarily unsupervised experiences and represented as abstract concepts. If such representations are compositional and hierarchical, they can be recombined into an exponentially large set of new concepts. This paper describes SCAN (Symbol-Concept Association Network), a new framework for learning such abstractions in the visual domain. SCAN learns concepts through fast symbol association, grounding them in disentangled visual primitives that are discovered in an unsupervised manner. Unlike state of the art multimodal generative model baselines, our approach requires very few pairings between symbols and images and makes no assumptions about the form of symbol representations. Once trained, SCAN is capable of multimodal bi-directional inference, generating a diverse set of image samples from symbolic descriptions and vice versa. It also allows for traversal and manipulation of the implicit hierarchy of visual concepts through symbolic instructions and learnt logical recombination operations. Such manipulations enable SCAN to break away from its training data distribution and imagine novel visual concepts through symbolically instructed recombination of previously learnt concepts.
| Type: | Proceedings paper |
|---|---|
| Title: | SCAN: Learning Hierarchical Compositional Visual Concepts |
| Event: | Sixth International Conference on Learning Representations (ICLR 2018) |
| Open access status: | An open access version is available from UCL Discovery |
| Publisher version: | https://openreview.net/forum?id=rkN2Il-RZ |
| Language: | English |
| Additional information: | This version is the version of record. For information on re-use, please refer to the publisher’s terms and conditions. |
| Keywords: | grounded visual concepts, compositional representation, concept hierarchy, disentangling, beta-VAE, variational autoencoder, deep learning, generative model |
| UCL classification: | UCL UCL > Provost and Vice Provost Offices > School of Life and Medical Sciences UCL > Provost and Vice Provost Offices > School of Life and Medical Sciences > Faculty of Life Sciences UCL > Provost and Vice Provost Offices > School of Life and Medical Sciences > Faculty of Life Sciences > Gatsby Computational Neurosci Unit UCL > Provost and Vice Provost Offices > UCL BEAMS UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Engineering Science UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Engineering Science > Dept of Computer Science |
| URI: | https://discovery.ucl.ac.uk/id/eprint/10124080 |
Archive Staff Only
 |
View Item |
