Privacy-Constrained Policies via Mutual Information Regularized Policy Gradients.
Chris CundyStefano ErmonPublished in: CoRR (2020)
Keyphrases
- mutual information
- privacy policies
- optimal policy
- privacy protection
- information theoretic
- personal information
- image registration
- similarity measure
- control policies
- management policies
- markov decision process
- private information
- feature selection
- access control policies
- third party
- privacy concerns
- privacy preserving
- medical image registration
- allocation policy
- transport systems
- personal data
- information gain
- role based access control
- policy search
- revenue management
- reward function
- access control
- markov decision processes
- infinite horizon
- multimodal image registration
- allocation policies
- conditional mutual information
- state space
- information theoretic measures
- markov decision problems
- policy gradient methods
- mutual information maximization
- reinforcement learning
- dynamic programming
- scheduling policies
- finite horizon
- least squares
- computer vision
- long run
- statistical databases
- average reward
- control policy
- partially observable markov decision processes
- security issues
- selective perception
- policy enforcement