Self Punishment and Reward Backfill for Deep Q-Learning.
Mohammad Reza BonyadiRui WangMaryam ZiaeiPublished in: CoRR (2020)
Keyphrases
- reinforcement learning
- agent receives
- reward function
- eligibility traces
- learning agent
- reinforcement learning algorithms
- multi agent
- cooperative
- function approximation
- state space
- discounted reward
- state action
- markov decision processes
- learning algorithm
- deep learning
- optimal policy
- temporal difference learning
- reinforcement learning methods
- dynamic programming
- hierarchical reinforcement learning
- model free
- machine learning
- multi agent reinforcement learning
- multi armed bandit
- continuous state spaces
- bandit problems
- continuous state and action spaces
- data sets
- policy gradient
- average reward
- long run
- evaluation function
- decision problems
- monte carlo
- least squares
- mobile robot
- neural network