Avoiding Tampering Incentives in Deep RL via Decoupled Approval.
Jonathan UesatoRamana KumarVictoria KrakovnaTom EverittRichard NgoShane LeggPublished in: CoRR (2020)
Keyphrases
- reinforcement learning
- digital images
- reverse engineering
- markov decision processes
- jpeg images
- deep learning
- optimal policy
- learning process
- multi agent
- machine learning
- watermarking scheme
- authentication scheme
- tamper detection
- signal processing
- supervised learning
- wavelet transform
- action selection
- jpeg compression
- moral hazard