Curriculum Learning for Motor Skills

By Andrej Karpathy, Michiel van de Panne. 2012

60 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Control and Systems Engineering, Reinforcement Learning, Human-Computer Interaction, Robotics & Control

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

1. Sergey Levine
2. Luc Van Gool
3. Trevor Darrell
4. Jitendra Malik
5. Andrea Vedaldi
6. Dacheng Tao
7. Xiaogang Wang
8. Ming-Hsuan Yang

In Control and Systems Engineering

1. Sergey Levine
2. Pieter Abbeel
3. David Held
4. Daniel Jaymin Mankowitz
5. Carlos Florensa
6. Kenneth O. Stanley
7. Jeff Clune
8. Markus Wulfmeier

In Reinforcement Learning

1. Sergey Levine
2. Pieter Abbeel
3. Jeff Clune
4. David Silver
5. Matteo Hessel
6. Chelsea Finn
7. Yuxi Li
8. Hado van Hasselt

In Human-Computer Interaction

1. Carlos Florensa
2. David Held
3. Pieter Abbeel
4. Xinyang Geng

In Robotics & Control

1. Carlos Florensa
2. David Held
3. David Silver
4. Pieter Abbeel
5. Tom Schaul
6. Kenneth O. Stanley
7. Michael Zhang
8. Jeff Clune

Citations in the corpus

• Reverse Curriculum Generation for Reinforcement Learning (2017, methodology): “This type of adaptive curriculum has mainly been applied for supervised tasks and most practical curriculum approaches in RL rely on pre-specified task sequences [14].”
• Paired Open-Ended Trailblazer (POET): Endlessly Generating Increasingly Complex and Diverse Learning Environments and Their Solutions (2019, background): “This kind of incremental curriculum is intuitive and variants of it appear in the literature when a task is too hard to learn directly [40, 67–70].”
• Enhanced POET: Open-Ended Reinforcement Learning through Unbounded Invention of Learning Challenges and their Solutions (2020, background): “Interestingly, much research effort is spent attempting to design or learn single-path curricula to help an agent learn a complex task (Bengio et al., 2009; Gomez & Miikkulainen, 1997; Heess et al., 2017; Justesen et al., 2018; Karpathy & Van De Panne, 2012).”
• Automatic Goal Generation for Reinforcement Learning Agents (2017, methodology): “However this has mainly been applied for supervised tasks and most curriculum learning in reinforcement learning still relies on fixed pre-specified sequences of tasks (Karpathy & Van De Panne, 2012).”
• Unicorn: Continual Learning with a Universal, Off-policy Agent (2018, background): “In particular, an automatic curriculum [22, 14, 15, 40, 36] with progressively more emphasis on harder tasks can accelerate learning via more efficient exploration and better generalization [7, 13, 14].”