Geometric deep learning on graphs and manifolds using mixture model CNNs
- Submitting institution
-
Imperial College of Science, Technology and Medicine
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 2415
- Type
- E - Conference contribution
- DOI
-
10.1109/cvpr.2017.576
- Title of conference / published proceedings
- 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
- First page
- 5425
- Volume
- 2017-January
- Issue
- -
- ISSN
- 1063-6919
- Open access status
- Deposit exception
- Month of publication
- November
- Year of publication
- 2017
- URL
-
-
- Supplementary information
-
10.1109/CVPR.2017.576
- Request cross-referral to
- -
- Output has been delayed by COVID-19
- No
- COVID-19 affected output statement
- -
- Forensic science
- No
- Criminology
- No
- Interdisciplinary
- No
- Number of additional authors
-
5
- Research group(s)
-
-
- Citation count
- 81
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- This unified framework generalises CNN architectures and outperforms other approaches for network analysis. The algorithm was patented (US15006694) leading to spinoff FabulaAI, acquired by Twitter, that performs fake news detection in social networks (accuracy >93% ROC AUC). Paper accepted for oral presentation at CVPR 2017 (acceptance rate: 2.65%/783) and resulted in multiple keynote invitations, including Graph Signal Processing 2018. Key concepts in the paper underpin Bronstein's ERC Consolidator grant 2016 (LEMAN; €2M), ERC Proof of Concept grant 2018 (GoodNews; €150K), Amazon AWS ML Research grant, Royal Society Wolfson Research Merit Award and Dalle Molle prize (major Swiss innovation award).
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -
