Python Reinforcement Learning
上QQ阅读APP看书,第一时间看更新

Dynamic programming

Dynamic programming (DP) is a technique for solving complex problems. In DP, instead of solving complex problems one at a time, we break the problem into simple sub-problems, then for each sub-problem, we compute and store the solution. If the same sub-problem occurs, we will not recompute, instead, we use the already computed solution. Thus, DP helps in drastically minimizing the computation time. It has its applications in a wide variety of fields including computer science, mathematics, bioinformatics, and so on.

We solve a Bellman equation using two powerful algorithms:

  • Value iteration
  • Policy iteration