Discovering modular solutions that generalize compositionally
This addresses the challenge of making AI systems more flexible and generalizable for complex tasks by uncovering conditions for modular discovery, though it is incremental as it builds on existing modular and meta-learning frameworks.
The paper tackles the problem of enabling compositional generalization in modular systems by studying a teacher-student setting with controlled ground truth modules, showing theoretically that identification up to linear transformation is possible without exponential combinations and empirically that meta-learning can discover modular policies that generalize compositionally in complex environments.
Many complex tasks can be decomposed into simpler, independent parts. Discovering such underlying compositional structure has the potential to enable compositional generalization. Despite progress, our most powerful systems struggle to compose flexibly. It therefore seems natural to make models more modular to help capture the compositional nature of many tasks. However, it is unclear under which circumstances modular systems can discover hidden compositional structure. To shed light on this question, we study a teacher-student setting with a modular teacher where we have full control over the composition of ground truth modules. This allows us to relate the problem of compositional generalization to that of identification of the underlying modules. In particular we study modularity in hypernetworks representing a general class of multiplicative interactions. We show theoretically that identification up to linear transformation purely from demonstrations is possible without having to learn an exponential number of module combinations. We further demonstrate empirically that under the theoretically identified conditions, meta-learning from finite data can discover modular policies that generalize compositionally in a number of complex environments.