Joseph Marino
Papers
1
Total Citations
7
H-Index
1
About
Joseph Marino is a researcher at the forefront of deep reinforcement learning (RL), with a particular focus on improving the stability and efficiency of value-based methods. His most-cited work, "Beyond Target Networks: Improving Deep Q-learning with Functional Regularization" (2021, 7 citations), critically examines the widely-used target network mechanism in deep Q-learning. Marino identifies a key limitation: while target networks stabilize training by using frozen parameters to estimate Q-values, they inadvertently slow down the propagation of new reward information. To address this, he introduces a functional regularization approach that retains stability without sacrificing learning speed, offering a principled alternative to standard target networks. This contribution is especially valuable for researchers tackling sample efficiency and convergence in complex RL environments. Marino’s work sits at the intersection of algorithmic theory and practical performance, making him a notable voice in the ongoing evolution of deep RL algorithms.
Research Focus
Key Achievements
Top Papers
- 1