Online Multi-Task Learning Using Active Sampling
By Balaraman Ravindran. 2017
21 citations recorded by the source corpus.
One of the long-standing challenges in Artificial Intelligence for goal-directed behavior is to build a single agent which can solve multiple tasks. Recent progress in multi-task learning for goal-directed sequential tasks has been in the form of distillation based learning wherein a single student network learns from multiple task-specific expert networks by mimicking the task-specific policies of the expert networks. While such approaches offer a promising solution to the multi-task learning problem, they require supervision from large task-specific (expert) networks which require extensive training. We propose a simple yet efficient multi-task learning framework which solves multiple goal-directed tasks in an online or active learning setup without the need for expert supervision.
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
- Sergey Levine
- Chelsea Finn
- Mohit Bansal
- Pieter Abbeel
- Tianhe Yu
- Yoshua Bengio
- Ruslan Salakhutdinov
- Trevor Darrell
Citations in the corpus
- A Survey on Multi-Task Learning (2017, background): “In [156], [157], multiple goaldirected tasks are learned in an online setup without the need for expert supervision by actively sampling harder tasks.”
- Meta-World: A Benchmark and Evaluation for Multi-Task and Meta Reinforcement Learning (2019, background): “While having a tremendous impact on the multi-task reinforcement learning research community [9, 15, 8, 16, 17], the Atari games included in the benchmark have significant differences in visual appearance, controls, and objectives, making it challenging to acquire any efficiency gains through shared”
- Multi-task Deep Reinforcement Learning with PopArt (2018, background): “2017), and parallel learning of multiple tasks at once (Sharma and Ravindran 2017; Caruana 1998).”
- AutoSeM: Automatic Task Selection and Mixing in Multi-Task Learning (2019, background): “Recently, Sharma and Ravindran (2017) used multi-armed bandits (MAB) to learn the choice of hard vs. easy domain data selection as input feed for the model.”
- Online Adaptative Curriculum Learning for GANs (2018, unknown): “s. Zhou and Bilmes (2018) designed a growing training set and implicitly learned a curriculum with the help of a submodular function promoting diversity in the data. In a multi-task learning context, Sharma and Ravindran (2017) used a Multi-Armed Bandit (MAB) algorithm to select which tasks the agen”