Principled Analysis of Deep Reinforcement Learning Evaluation and Design Paradigms
For the deep reinforcement learning research community, this work challenges the validity of established evaluation and design practices, potentially reshaping how algorithms are compared and developed.
The paper analyzes canonical evaluation and design paradigms in deep reinforcement learning, introducing theoretical foundations of scaling laws and showing that asymptotic performance rankings are not monotonic with data regimes. Large-scale experiments reveal that following these paradigms has led to incorrect conclusions in prior research.
Starting from the utilization of deep neural networks to approximate the state-action value function that led to winning one of the most challenging games, to algorithmic advancements that allowed solving problems without even explicitly stating the rules of the challenge at hand, reinforcement learning research has been the center of remarkable scientific progress for the past decade. In this paper, we focus on the key ingredients of this research progress and we analyze the canonical evaluation and design paradigms in reinforcement learning. We introduce the theoretical foundations of scaling laws in reinforcement learning and show that the asymptotic performance of reinforcement learning algorithms does not have a monotone relationship between performance rankings and data-regimes. We conduct large-scale experiments and our results demonstrate that a line of reinforcement learning research under the canonical design and evaluation paradigms resulted in incorrect conclusions. Our analysis and results provide a core analysis on scaling, capacity and complexity of deep reinforcement learning.