Two Steps Reinforcement Learning in Continuous Reinforcement Learning Tasks
نویسندگان
چکیده
Two steps reinforcement learning is a technique that combines an iterative refinement of a Q function estimator that can be used to obtains a state space discretization with classical reinforcement learning algorithms like Q-learning or Sarsa. However, the method requires a discrete reward function that permits learning an approximation of the Q function using classification algorithms. However, many domains have continuous reward functions that could only be tackled by discretizing the rewards. In this paper we propose solutions to this problem using discretization and regression methods. We demonstrate the usefulness of the resulting approach to improve the learning process in the Keepaway domain. We compare the obtained results with other techniques like VQQL and CMAC.
منابع مشابه
Reinforcement Learning in Neural Networks: A Survey
In recent years, researches on reinforcement learning (RL) have focused on bridging the gap between adaptive optimal control and bio-inspired learning techniques. Neural network reinforcement learning (NNRL) is among the most popular algorithms in the RL framework. The advantage of using neural networks enables the RL to search for optimal policies more efficiently in several real-life applicat...
متن کاملReinforcement Learning in Neural Networks: A Survey
In recent years, researches on reinforcement learning (RL) have focused on bridging the gap between adaptive optimal control and bio-inspired learning techniques. Neural network reinforcement learning (NNRL) is among the most popular algorithms in the RL framework. The advantage of using neural networks enables the RL to search for optimal policies more efficiently in several real-life applicat...
متن کاملEfficient Continuous-Time Reinforcement Learning with Adaptive State Graphs
We present a new reinforcement learning approach for deterministic continuous control problems in environments with unknown, arbitrary reward functions. The difficulty of finding solution trajectories for such problems can be reduced by incorporating limited prior knowledge of the approximative local system dynamics. The presented algorithm builds an adaptive state graph of sample points within...
متن کاملPractical Reinforcement Learning in Continuous Spaces
Dynamic control tasks are good candidates for the application of reinforcement learning techniques. However, many of these tasks inherently have continuous state or action variables. This can cause problems for traditional reinforcement learning algorithms which assume discrete states and actions. In this paper, we introduce an algorithm that safely approximates the value function for continuou...
متن کاملGeneralization in Reinforcement Learning
In this paper we evaluate two Temporal Difference Reinforcement Learning methods on several different tasks to see how well these methods generalize. The tasks were modeled as Markov Decision Processes with a continuous observation space and a discrete action space. Function approximation was done using linear gradient descent with RBFs as features. The tasks were taken from the Polyathlon doma...
متن کامل