Trial without Error: Towards Safe Reinforcement Learning via Human Intervention.
William SaundersGirish SastryAndreas StuhlmüllerOwain EvansPublished in: AAMAS (2018)
Keyphrases
- reinforcement learning
- error rate
- error bounds
- linear complexity
- model free
- function approximation
- dynamic programming
- markov decision processes
- action selection
- error analysis
- transfer learning
- state space
- multi agent
- learning algorithm
- machine learning
- neural network
- data sets
- estimation error
- knowledge base
- temporal difference
- reinforcement learning algorithms
- learning capabilities
- approximation error