Generalizing to New Tasks via One-Shot Compositional Subgoals
This work addresses the problem of generalization to unseen tasks with minimal supervision for AI agents in real-world applications, representing a strong specific gain in imitation learning.
The paper tackles the challenge of sample inefficiency and long training times in long-horizon tasks by introducing CASE, an imitation learning method that uses adaptive near-future subgoals in a learned latent space, achieving a 30% improvement over the previous state-of-the-art compositional imitation learning approach.
The ability to generalize to previously unseen tasks with little to no supervision is a key challenge in modern machine learning research. It is also a cornerstone of a future "General AI". Any artificially intelligent agent deployed in a real world application, must adapt on the fly to unknown environments. Researchers often rely on reinforcement and imitation learning to provide online adaptation to new tasks, through trial and error learning. However, this can be challenging for complex tasks which require many timesteps or large numbers of subtasks to complete. These "long horizon" tasks suffer from sample inefficiency and can require extremely long training times before the agent can learn to perform the necessary longterm planning. In this work, we introduce CASE which attempts to address these issues by training an Imitation Learning agent using adaptive "near future" subgoals. These subgoals are recalculated at each step using compositional arithmetic in a learned latent representation space. In addition to improving learning efficiency for standard long-term tasks, this approach also makes it possible to perform one-shot generalization to previously unseen tasks, given only a single reference trajectory for the task in a different environment. Our experiments show that the proposed approach consistently outperforms the previous state-of-the-art compositional Imitation Learning approach by 30%.