Dor Livne

Papers

Total Citations

H-Index

About

Dor Livne is a leading researcher at the intersection of deep reinforcement learning (DRL) and model efficiency, with a primary focus on making DRL systems more practical for real-world deployment. His most impactful work, "PoPS: Policy Pruning and Shrinking for Deep Reinforcement Learning" (2020, 41 citations), introduces a novel framework that systematically compresses deep neural network policies without sacrificing performance. By combining structured pruning with knowledge distillation, Livne demonstrated that DRL agents could operate with significantly reduced computational and memory footprints—a critical advancement for resource-constrained applications in robotics, embedded systems, and edge computing. This contribution addresses a fundamental bottleneck in DRL, where overparameterized networks often hinder deployment in latency-sensitive or low-power environments. Livne’s research has been recognized for bridging the gap between theoretical algorithm design and practical engineering constraints, earning him citations across fields like autonomous navigation and game AI. His work continues to influence the development of lean, efficient reinforcement learning systems that maintain high performance while enabling scalable, real-time decision-making in dynamic environments.