MaxGNR: A Dynamic Weight Strategy via Maximizing Gradient-to-Noise Ratio for Multi-task Learning.
Caoyun FanWenqing ChenJidong TianYitian LiHao HeYaohui JinPublished in: ACCV (1) (2022)
Keyphrases
- multi task learning
- multi task
- learning tasks
- multiple tasks
- gaussian processes
- multitask learning
- learning problems
- high order
- multi label image annotation
- transfer learning
- noise ratio
- computer simulation
- image classification
- higher order
- active learning
- prior knowledge
- bayesian networks
- decision trees
- social networks
- learning algorithm