Discovering symbolic policies with deep reinforcement learning.
Mikel Landajuela LarmaBrenden K. PetersenSookyung KimCláudio P. SantiagoRuben GlattT. Nathan MundhenkJacob F. PettitDaniel FaissolPublished in: ICML (2021)
Keyphrases
- reinforcement learning
- optimal policy
- policy search
- markov decision process
- reinforcement learning agents
- control policies
- markov decision processes
- reward function
- function approximation
- cooperative multi agent systems
- markov decision problems
- reinforcement learning algorithms
- policy gradient methods
- fitted q iteration
- state space
- total reward
- partially observable markov decision processes
- hierarchical reinforcement learning
- learning algorithm
- model free
- macro actions
- finite state
- control policy
- policy iteration
- temporal difference
- symbolic representation
- machine learning
- infinite horizon
- decision problems
- transfer learning
- deep learning
- average reward
- multi agent
- reward shaping
- dynamic programming
- high level
- robotic control
- tabula rasa
- learning process
- supervised learning
- multiagent reinforcement learning
- multi agent reinforcement learning
- continuous state
- temporal difference learning
- function approximators