White-Box Adversarial Policies in Deep Reinforcement Learning.
Stephen CasperDylan Hadfield-MenellGabriel KreimanPublished in: SafeAI@AAAI (2023)
Keyphrases
- white box
- reinforcement learning
- black box
- optimal policy
- policy search
- markov decision process
- source code
- multi agent
- control policies
- markov decision processes
- reinforcement learning agents
- fitted q iteration
- reward function
- reinforcement learning algorithms
- function approximation
- state space
- hierarchical reinforcement learning
- control policy
- markov decision problems
- decision problems
- continuous state
- model free
- test cases
- transfer learning
- test data
- learning process
- reinforcement learning methods
- machine learning
- partially observable markov decision processes
- partially observable
- case study
- infinite horizon
- source code metrics
- decentralized control
- function approximators
- long run
- finite state
- average reward
- action selection
- actor critic
- open source
- multiagent reinforcement learning
- supervised learning
- relational databases
- learning algorithm
- neural network
- databases