HYPERDQN: A RANDOMIZED EXPLORATION METHOD FOR DEEP REINFORCEMENT LEARNING

Keyphrases

• Value Function 100%
• Exploration Method 100%
• Deep Reinforcement Learning (deep RL) 100%
• Value Iteration 75%
• Randomized Least Squares 75%
• Least Square Value 75%
• Posterior Sampling 75%
• Efficient Exploration 50%
• Good Characteristics 50%
• Hypermodel 50%
• Nonlinear Neural Networks 25%
• Computational Burden 25%
• Practical Algorithm 25%
• Metamodel 25%
• Approximate Posterior 25%
• Atari 25%
• Exploration Bonus 25%
• Exploratory Action 25%
• Action Sequences 25%
• Jointly Optimization 25%
• Normalized Score 25%

Mathematics

• Posterior Sample 100%
• Least Squares Method 100%
• Deep Reinforcement Learning 100%
• Approximates 66%
• Exploratory 33%
• Neural Network 33%
• Nonlinear 33%

Computer Science

• Least Squares Method 100%
• Deep Reinforcement Learning 100%
• Neural Network 33%
• Action Sequence 33%