Parameter Space Noise for Exploration
This work addresses exploration inefficiencies in reinforcement learning for agents, offering a method that combines the benefits of parameter perturbations with temporal structure, though it is incremental as it builds on existing RL techniques.
The paper tackled the problem of exploration in deep reinforcement learning by proposing parameter space noise as an alternative to action space noise, resulting in more efficient learning across high-dimensional discrete and continuous control tasks compared to traditional RL and evolutionary strategies.
Deep reinforcement learning (RL) methods generally engage in exploratory behavior through noise injection in the action space. An alternative is to add noise directly to the agent's parameters, which can lead to more consistent exploration and a richer set of behaviors. Methods such as evolutionary strategies use parameter perturbations, but discard all temporal structure in the process and require significantly more samples. Combining parameter noise with traditional RL methods allows to combine the best of both worlds. We demonstrate that both off- and on-policy methods benefit from this approach through experimental comparison of DQN, DDPG, and TRPO on high-dimensional discrete action environments as well as continuous control tasks. Our results show that RL with parameter noise learns more efficiently than traditional RL with action space noise and evolutionary strategies individually.