Hypernetworks That Evolve Themselves
This work addresses the challenge of autonomous, open-ended learning for AI agents, representing an incremental step toward synthetic systems that mimic biological evolution.
The paper tackled the problem of enabling neural networks to evolve autonomously without external optimizers by proposing Self-Referential Graph HyperNetworks, which showed swift adaptation in reinforcement learning benchmarks like CartPoleSwitch and evolved coherent gaits in Ant-v5.
How can neural networks evolve themselves without relying on external optimizers? We propose Self-Referential Graph HyperNetworks, systems where the very machinery of variation and inheritance is embedded within the network. By uniting hypernetworks, stochastic parameter generation, and graph-based representations, Self-Referential GHNs mutate and evaluate themselves while adapting mutation rates as selectable traits. Through new reinforcement learning benchmarks with environmental shifts (CartPoleSwitch, LunarLander-Switch), Self-Referential GHNs show swift, reliable adaptation and emergent population dynamics. In the locomotion benchmark Ant-v5, they evolve coherent gaits, showing promising fine-tuning capabilities by autonomously decreasing variation in the population to concentrate around promising solutions. Our findings support the idea that evolvability itself can emerge from neural self-reference. Self-Referential GHNs reflect a step toward synthetic systems that more closely mirror biological evolution, offering tools for autonomous, open-ended learning agents.