Hereditary Geometric Meta-RL: Nonlocal Generalization via Task Symmetries

Researchers developed a new Meta-Reinforcement Learning approach that uses geometric symmetries in task spaces to enable broader generalization beyond local smoothness assumptions. The method converts Meta-RL into symmetry discovery rather than smooth extrapolation, allowing agents to generalize across wider regions of task space with improved sample efficiency.

• →New geometric Meta-RL approach uses task symmetries instead of smoothness for generalization.
• →Method converts Meta-RL from smooth extrapolation to symmetry discovery problem.
• →Task space embeds into linearizable, connected, and compact subgroups enabling efficient learning.
• →Differential symmetry discovery method improves numerical stability and sample efficiency.
• →Empirical results show full task space generalization compared to baseline methods that only work near training tasks.