Pathologies of temporal difference methods in approximate dynamic programming

Mathematics

• Approximate Dynamic Programming 100%
• Difference Method 71%
• Policy Iteration 59%
• Reinforcement Learning 30%
• Testbed 29%
• Random Search 27%
• Search Methods 24%
• Experimentation 24%
• Gradient Method 22%
• Local Minima 22%
• Iteration Method 21%
• Likely 20%
• Policy 20%
• Game 16%
• Oscillation 16%
• Standards 11%

Engineering & Materials Science

• Gradient methods 74%
• Pathology 69%
• Reinforcement learning 61%
• Dynamic programming 59%
• Testbeds 59%