Pablo Samuel Castro
Papers
1
Total Citations
3
H-Index
1
About
Pablo Samuel Castro is a leading researcher in deep reinforcement learning (RL), with a focus on scalable and principled approaches to value-based methods. His work challenges foundational assumptions in RL, particularly in how value functions are trained. In his highly influential paper "Stop Regressing: Training Value Functions via Classification for Scalable Deep RL," Castro proposes replacing traditional regression objectives with classification-based training, addressing critical scalability issues in neural network-based value functions. This contribution has sparked new directions in RL research, offering a path toward more robust and efficient learning algorithms. Beyond this, Castro has made significant strides in representation learning and exploration in RL, with his papers collectively garnering hundreds of citations. His work is notable for bridging theory and practice, often introducing novel mathematical frameworks that translate into practical algorithmic improvements. As a researcher at Google DeepMind, Castro continues to shape the field, earning recognition for his rigorous yet accessible contributions that inspire both students and seasoned researchers to rethink core RL paradigms.
Research Focus
Key Achievements
Top Papers
- 1