Rethinking Population-Assisted Off-policy Reinforcement Learning

Keyphrases

• Policy Learning 100%
• Replay Buffer 60%
• Population Data 40%
• Population-based Algorithms 40%
• Hybrid Method 20%
• Local Optimum 20%
• Black Box 20%
• Gradient-based 20%
• Data Reuse 20%
• Training Design 20%
• Exploration Strategy 20%
• Buffer Design 20%
• Robot Locomotion 20%
• Diverse Data 20%
• Optimization Iteration 20%
• On-policy 20%
• Scalable Training 20%
• Population-based Optimization 20%
• OpenAI Gym 20%
• Sample Efficiency 20%
• Locomotion Tasks 20%

Computer Science

• Reinforcement Learning 100%
• Data Reuse 20%
• Robot Locomotion 20%