Papers
10
Total Citations
912
H-Index
9
About
Thomas Degris is a prominent researcher at the intersection of reinforcement learning, robotics, and adaptive human-machine interaction, whose work has shaped how intelligent systems learn and operate in real-world environments. He is perhaps best known for co-developing **Horde**, a scalable real-time architecture that enables robots to build rich world knowledge through unsupervised sensorimotor interaction using parallel reinforcement learning sub-agents called "demons" — a landmark contribution that has accumulated over 300 citations and influenced subsequent work in continual and multi-task learning. Degris has made substantial contributions to practical reinforcement learning with continuous actions, addressing the critical challenge of real-time robot adaptability, a paper that has garnered over 230 citations. His work on myoelectric prosthesis control stands out as particularly impactful, applying actor-critic reinforcement learning to help amputees more effectively operate adaptive artificial limbs, earning over 150 citations and demonstrating the real human stakes of his research. He has also advanced fundamental algorithmic challenges, including step-size adaptation and meta-descent methods for continual learning. Across his career, Degris has consistently bridged theoretical machine learning with meaningful applications in assistive robotics and intelligent systems, making his work valuable reading for researchers in both fields.
Research Focus
Key Achievements
Top Papers
- 1
- 2Model-Free reinforcement learning with continuous action in practice234 citations · 2012
- 3
- 4
- 5
- 6Tuning-free step-size adaptation52 citations · 2012
- 7A Spiking Neuron Model of Head-Direction Cells for Robot Orientation19 citations · 2004
- 8Scaling-up Knowledge for a Cognizant Robot12 citations · 2012
- 9Meta-Descent for Online, Continual Prediction11 citations · 2019
- 10