Playing 20 Question Game with Policy-Based Reinforcement Learning.
Huang HuXianchao WuBingfeng LuoChongyang TaoCan XuWei WuZhan ChenPublished in: EMNLP (2018)
Keyphrases
- reinforcement learning
- game playing
- computer games
- optimal policy
- policy search
- agent learns
- temporal difference learning
- imperfect information
- markov decision process
- action selection
- game players
- card game
- markov games
- pac man
- board game
- multi player
- reinforcement learning algorithms
- function approximation
- online game
- games played
- markov decision processes
- function approximators
- partially observable environments
- educational games
- reward function
- game play
- policy gradient
- human players
- reinforcement learning problems
- state and action spaces
- game design
- game theory
- video games
- action space
- rl algorithms
- policy evaluation
- state space
- markov decision problems
- control policies
- state action
- policy iteration
- game tree search
- virtual world
- actor critic
- temporal difference
- average reward
- game theoretic
- partially observable
- general game playing
- inverse reinforcement learning
- learning algorithm
- perfect information
- model free
- control policy
- two player games
- multi agent
- minimax search
- partially observable domains
- dynamic programming
- computer programs
- game tree
- partially observable markov decision processes
- continuous state
- machine learning
- serious games
- continuous state spaces
- decision problems
- stochastic games
- reinforcement learning methods
- nash equilibrium
- infinite horizon