| Sign In to gain access to subscriptions and/or personal tools. |
Co-evolution of Shaping Rewards and Meta-Parameters in Reinforcement LearningCentre for Autonomous Systems, Numerical Analysis and Computer Science, KTH, Sweden, Neural Computation Unit, Okinawa Institute of Science and Technology, Japan, elfwing{at}oist.jp
Neural Computation Unit, Okinawa Institute of Science and Technology, Japan, uchibe{at}oist.jp
Neural Computation Unit, Okinawa Institute of Science and Technology, Japan, doya{at}oist.jp
Centre for Autonomous Systems, Numerical Analysis and Computer Science, KTH, Sweden, hic{at}kth.se
In this article, we explore an evolutionary approach to the optimization of potential-based shaping rewards and meta-parameters in reinforcement learning. Shaping rewards is a frequently used approach to increase the learning performance of reinforcement learning, with regards to both initial performance and convergence speed. Shaping rewards provide additional knowledge to the agent in the form of richer reward signals, which guide learning to high-rewarding states. Reinforcement learning depends critically on a few meta-parameters that modulate the learning updates or the exploration of the environment, such as the learning rate
Key Words: reinforcement learning • shaping rewards • meta-parameters • genetic algorithms
Adaptive Behavior, Vol. 16, No. 6,
400-412 (2008) |
|||
 |
|
|
 |
|
Sign In to gain access to subscriptions and/or personal tools.
, the discount factor of future rewards 
, and the temperature 
that controls the trade-off between exploration and exploitation in softmax action selection. We validate the proposed approach in simulation using the mountain-car task. We also transfer shaping rewards and meta-parameters, evolutionarily obtained in simulation, to hardware, using a robotic foraging task.