(Georgia Institute of Technology, 2013)
Cobo, Luis C.; Isbell, Charles L.; Thomaz, Andrea L.
We present Object Focused Q-learning (OF-Q), a novel reinforcement learning algorithm that can offer exponential speed-ups over classic Q-learning on domains composed of independent objects. An OF-Q agent treats the state space as a collection of objects organized into different object classes. Our key contribution is a control policy that uses non-optimal Q-functions to estimate the risk of ignoring parts of the state space. We compare our algorithm to traditional Q-learning and previous arbitration algorithms in two domains, including a version of Space Invaders.
(Georgia Institute of Technology, 2012-06)
Cobo, Luis C.; Isbell, Charles L.; Thomaz, Andrea L.
Both Learning from Demonstration (LfD) and Reinforcement Learning (RL) are popular approaches for building decision-making agents. LfD applies supervised learning to a set of human demonstrations to infer and imitate the human policy, while RL uses only a reward signal and exploration to find an optimal policy. For complex tasks both of these techniques may be ineffective. LfD may require many more demonstrations than it is feasible to obtain, and RL can take an inadmissible amount of time to converge. We present Automatic Decomposition and Abstraction from demonstration (ADA), an algorithm that uses mutual information measures over a set of human demonstrations to decompose a sequential decision process into several sub- tasks, finding state abstractions for each one of these sub- tasks. ADA then projects the human demonstrations into the abstracted state space to build a policy. This policy can later be improved using RL algorithms to surpass the performance of the human teacher. We find empirically that ADA can find satisficing policies for problems that are too complex to be solved with traditional LfD and RL algorithms. In particular, we show that we can use mutual information across state features to leverage human demonstrations to reduce the effects of the curse of dimensionality by finding subtasks and abstractions in sequential decision processes.