Consistent Multitask Learning with Nonlinear Output Relations.
- Submitting institution
-
University College London
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 14277
- Type
- E - Conference contribution
- DOI
-
-
- Title of conference / published proceedings
- Advances in Neural Information Processing Systems 30 (NIPS 2017)
- First page
- 1983
- Volume
- 2017-December
- Issue
- -
- ISSN
- 1049-5258
- Open access status
- Compliant
- Month of publication
- December
- Year of publication
- 2017
- URL
-
-
- Supplementary information
-
-
- 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
-
3
- Research group(s)
-
-
- Citation count
- 0
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- Originally: We propose a general framework for nonlinear multitask learning, based on which we derive a computationally efficiently and reliable algorithm. Significance: Previous methods on multitask learning (featured in over 1000 publications) have focused almost exclusively on modelling linear relationships between the tasks, which may be too restrictive in practice. Here we show that techniques from structure prediction are well-suited to study multitask learning problems that are related via a set of nonlinear output constraints. Rigour: Our theoretical analysis is based on sharp bounds that highlight the benefits of our multitask approach over standard multitask learning.
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -