pyrobolearn/pyrobolearn/tasks

Learning Tasks / paradigms

This folder contains learning tasks/paradigms and how they should be run. These include:

• Reinforcement learning (RL) tasks: the task accepts as inputs the environment and the policies. This can then be given to a RL algorithm.
• Imitation learning (IL) tasks: the task accepts as inputs the environment, the policies, and the bridges to the interface used to interact with a part of the simulator (robot, object in the world, etc).

Other learning paradigms include (but are not implemented yet):

• Active learning (AL)
• Curriculum learning (CL)
• Inverse reinforcement learning (IRL)
• Transfer learning (TL)
• Hierarchical learning (HL)
• Online learning (OL)
• Lifelong (or continual) learning (LL)
• Multi-task learning (MTL)
• Meta-learning (ML)

Each task/paradigm can be evaluated using different metrics.

TODO

• maybe I should change the name of the folder to paradigms instead of tasks to avoid the confusion with robotic tasks (used for instance in the stack of tasks).
• implement the other paradigms