Timezone: »

 
Model-Based Meta Automatic Curriculum Learning
Zifan Xu · Yulin Zhang · Shahaf Shperberg · Reuth Mirsky · Yuqian Jiang · Bo Liu · Peter Stone

@
in Decision Awareness in Reinforcement Learning »
Event URL: https://openreview.net/forum?id=Kp716SJ5dbJ »

When an agent trains for one target task, its experience is expected to be useful for training on another target task. This paper formulates the meta curriculum learning problem that builds a sequence of intermediate training tasks, called a curriculum, which will assist the learner to train toward any given target task in general. We propose a model-based meta automatic curriculum learning algorithm (MM-ACL) that learns to predict the performance on one task when trained on another, given contextual information such as the history of training tasks, loss functions, rollout state-action trajectories from the policy, etc. This predictor facilitates the generation of a curriculum that optimizes the performance of the learner on different target tasks. Our empirical results demonstrate that MM-ACL outperforms a random curriculum, a manually created curriculum, and a commonly used non-stationary bandit algorithm in a GridWorld domain.

Author Information

Zifan Xu (University of Texas at Austin)
Yulin Zhang (, University of Texas at Austin)
Shahaf Shperberg (University of Texas at Austin)
Reuth Mirsky (The University of Texas at Austin)
Yuqian Jiang (, University of Texas at Austin)
Bo Liu (University of Texas, Austin)
Peter Stone (The University of Texas at Austin and Sony AI)

More from the Same Authors