Please enable JavaScript.
Coggle requires JavaScript to display documents.
从数学角度分析方法的 Mathematical problem (关心的PROBLEM: (planning 忽略domain-specific…
从数学角度分析方法的
Mathematical problem
关心的METHOD：
dynamic programming
reinforcement learning
Optimal control
Multi-arm bandits
MDP
关心的PROBLEM:
提高search performance:
learning evaluation function
结合 prioritization, partitioning, and variance reordering
planning 忽略domain-specific knowledge
learning control knowledge for planning
Exploration in Relational Domains for model-based RL
Integrated Common Sense Learning and Planning in POMDPs
不同方法结合to improve performance
PI2: combine optimal control (path integral) with RL
combine tree method with RL
改进DP 方法
Approximate Modified Policy Iteration
safety learning
综述了： safe RL
combine Lyapunov method with RL
改进RL
Off-policy Learning With Eligibility Traces:
改进policy optimization
Approximate Newton Methods for Policy Search
针对高维state-action space的representation问题
regularization-based approximate value function
dynamic environment:
Value-Based RL in Changing Markovian Environments
综述类
探讨optimization and machine learning 关系