BiERL: A Meta Evolutionary Reinforcement Learning Framework via Bilevel Optimization
This addresses hyperparameter tuning challenges in ERL for researchers and practitioners, but it is incremental as it builds on existing ERL methods.
The paper tackles the problem of insufficient exploration and model collapse in evolutionary reinforcement learning (ERL) algorithms due to hyperparameter tuning, by proposing a meta ERL framework via bilevel optimization (BiERL) that jointly updates hyperparameters during training, and it outperforms baselines in MuJoCo and Box2D tasks.
Evolutionary reinforcement learning (ERL) algorithms recently raise attention in tackling complex reinforcement learning (RL) problems due to high parallelism, while they are prone to insufficient exploration or model collapse without carefully tuning hyperparameters (aka meta-parameters). In the paper, we propose a general meta ERL framework via bilevel optimization (BiERL) to jointly update hyperparameters in parallel to training the ERL model within a single agent, which relieves the need for prior domain knowledge or costly optimization procedure before model deployment. We design an elegant meta-level architecture that embeds the inner-level's evolving experience into an informative population representation and introduce a simple and feasible evaluation of the meta-level fitness function to facilitate learning efficiency. We perform extensive experiments in MuJoCo and Box2D tasks to verify that as a general framework, BiERL outperforms various baselines and consistently improves the learning performance for a diversity of ERL algorithms.