Kate Rakelly
Papers
1
Total Citations
3
H-Index
1
About
Kate Rakelly is a leading researcher at the intersection of meta-learning, reinforcement learning, and robotics, with a focus on enabling autonomous agents to adapt rapidly from high-dimensional sensory data. Her most influential work, "MELD: Meta-Reinforcement Learning from Images via Latent State Models," addresses a critical bottleneck in AI: the immense data demands of meta-training when learning directly from visual inputs. By introducing a method that learns compact latent state representations from images, Rakelly’s framework allows robots to efficiently acquire new skills by leveraging prior task experience—dramatically reducing the sample complexity required for real-world adaptation. This contribution has been foundational for advancing sample-efficient, vision-based meta-reinforcement learning, earning her work recognition as a key step toward practical, generalist robots. With over 3 citations on this seminal paper alone, Rakelly’s research continues to shape how machines learn to learn from raw perceptual streams, bridging the gap between simulation and physical deployment. Her achievements underscore a commitment to making autonomous systems more agile and data-efficient, inspiring new directions in lifelong learning and embodied AI.
Research Focus
Key Achievements
Top Papers
- 1MELD: Meta-Reinforcement Learning from Images via Latent State Models3 citations · 2020