Safer reinforcement learning through evolved instincts

Keyphrases

• Reinforcement Learning 100%
• Instinct 100%
• Hazard Zone 100%
• Safe Reinforcement Learning 100%
• New Targets 100%
• AI Safety 50%
• Action Space 50%
• Navigation Task 50%
• Meta-learning 50%
• Lifetime Learning 50%
• Inherent Risk 50%
• Noisy Output 50%
• 2D Navigation 50%
• State of the Environment 50%
• Deep Reinforcement Learning (deep RL) 50%

Computer Science

• Reinforcement Learning 100%
• Trained Network 66%
• Deep Reinforcement Learning 33%
• Navigation Task 33%
• Artificial Intelligence 33%