R-SQAIR: Relational Sequential Attend, Infer, Repeat

By Aleksandar Stani'c, J. Urgen Schmidhuber. 2019

33 citations recorded by the source corpus.

Traditional sequential multi-object attention models rely on a recurrent mechanism to infer object relations. We propose a relational extension (R-SQAIR) of one such attention model (SQAIR) by endowing it with a module with strong relational inductive bias that computes in parallel pairwise interactions between inferred objects. Two recently proposed relational modules are studied on tasks of unsupervised learning from videos. We demonstrate gains over sequential relational mechanisms, also in terms of combinatorial generalization.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Representation & Self-Supervised Learning, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Josh Tenenbaum
2. J. Urgen Schmidhuber
3. Sjoerd van Steenkiste
4. Li Fei-Fei
5. Jiajun Wu
6. Yoshua Bengio
7. Sergey Levine
8. Mohit Bansal

In Computer Vision and Pattern Recognition

1. Luc Van Gool
2. Li Fei-Fei
3. Josh Tenenbaum
4. Andrea Vedaldi
5. Jitendra Malik
6. Trevor Darrell
7. Dacheng Tao
8. Ming-Hsuan Yang

In Representation & Self-Supervised Learning

1. Sjoerd van Steenkiste
2. J. Urgen Schmidhuber
3. Jiajun Wu
4. Josh Tenenbaum
5. Aleksandar Stani'c
6. Li Fei-Fei
7. Daniel L. K. Yamins
8. Aran Nayebi

In Signal Processing

1. Li Fei-Fei
2. Sjoerd van Steenkiste
3. Aleksandar Stani'c
4. J. Urgen Schmidhuber
5. Josh Tenenbaum
6. Daniel L. K. Yamins
7. Jiajun Wu
8. Yoshua Bengio

Citations in the corpus

• Learning Physical Graph Representations from Visual Scenes (2020, background): “While most related works on inferring scene graph representations from visual data require fine-grained annotations, several works have studied inferring objects and their relations from videos without supervision [53, 46, 56].”
• Hierarchical Relational Inference (2020, methodology): “…representations from raw visual observations typically make use of mixture models (Greff, van Steenkiste, and Schmidhuber 2017; Greff et al. 2019) or attention (Eslami et al. 2016; Kosiorek et al. 2018; Stanić and Schmidhuber 2019; Burgess et al. 2019; Crawford and Pineau 2019; Jiang et al. 2020).”