FEWER MAY BE BETTER: ENHANCING OFFLINE REINFORCEMENT LEARNING WITH REDUCED DATASET

Keyphrases

• Reduced Data 100%
• Offline Reinforcement Learning 100%
• Algorithm Performance 40%
• Reinforcement Learning 40%
• Optimization Objectives 20%
• Optimization Problem 20%
• Novel Modification 20%
• Data Requirements 20%
• Limited Datasets 20%
• Low Computational Complexity 20%
• Performance Efficiency 20%
• Dataset Size 20%
• Algorithm Training 20%
• Similar Problem 20%
• Interaction with Environment 20%
• Subset Selection 20%
• Data Selection 20%
• Learning Research 20%
• Training Effectiveness 20%
• Orthogonal Matching Pursuit 20%
• Optimal Subset 20%
• Boosting Algorithm 20%
• Submodular Optimization 20%
• Gradient Approximation 20%
• Actor-critic Methods 20%
• Offline RL 20%
• Efficient Subset 20%
• Approximation Optimization 20%

Computer Science

• Reinforcement Learning 100%
• Algorithm Performance 28%
• Approximation (Algorithm) 14%
• Experimental Result 14%
• Optimisation Objective 14%
• Optimization Problem 14%
• Computational Complexity 14%
• Feature Subset Selection 14%
• Data Requirement 14%
• Orthogonal Matching Pursuit 14%