Mitigating Deep Reinforcement Learning Backdoors in the Neural Activation Space.
Sanyam VyasChris HicksVasilios MavroudisPublished in: SP (Workshops) (2024)
Keyphrases
- reinforcement learning
- network architecture
- function approximation
- machine learning
- information processing
- dynamic programming
- search space
- optimal policy
- parameter space
- satisfiability problem
- genetic algorithm
- biologically inspired
- data sets
- higher dimensional
- model free
- temporal difference
- deep learning
- function approximators
- robotic control
- lateral inhibition