Eliminating Meta Optimization Through Self-Referential Meta Learning
This addresses the need for automated learning algorithm design in AI, reducing human engineering overhead, though it appears incremental as it builds on existing meta learning concepts.
The paper tackles the problem of meta learning's dependency on human-designed meta optimization by proposing self-referential meta learning systems that modify themselves without explicit meta optimization, resulting in a neural network that self-modifies to solve tasks like bandit and classic control, improves its modifications, and learns how to learn by allocating more resources to better solutions.
Meta Learning automates the search for learning algorithms. At the same time, it creates a dependency on human engineering on the meta-level, where meta learning algorithms need to be designed. In this paper, we investigate self-referential meta learning systems that modify themselves without the need for explicit meta optimization. We discuss the relationship of such systems to in-context and memory-based meta learning and show that self-referential neural networks require functionality to be reused in the form of parameter sharing. Finally, we propose fitness monotonic execution (FME), a simple approach to avoid explicit meta optimization. A neural network self-modifies to solve bandit and classic control tasks, improves its self-modifications, and learns how to learn, purely by assigning more computational resources to better performing solutions.