Dumitru Erhan

Papers

Total Citations

H-Index

About

Dumitru Erhan is a leading researcher in deep learning, with a primary focus on generative models, reinforcement learning, and video understanding. His most influential work, "VideoFlow: A Flow-Based Generative Model for Video" (2019, 87 citations), introduced a novel approach to modeling and predicting video sequences using normalizing flows, enabling the capture of complex real-world phenomena such as physical interactions. This contribution advanced the field of generative video modeling by offering a principled framework for learning predictive models of dynamic visual scenes. Erhan has also explored model-based reinforcement learning, as seen in his work "INFOrmation Prioritization through EmPOWERment in Visual Model-Based RL" (2022), which addresses the challenge of learning latent state representations that prioritize functionally relevant information in complex visual environments. His research bridges generative modeling and reinforcement learning, with applications in robotics and autonomous systems. Erhan’s work is widely recognized for its technical depth and practical implications, making him a notable figure in the AI community. His contributions continue to inspire researchers working on video generation, representation learning, and decision-making under uncertainty.