• If problems are well formed, means-ends analysis and problem reduction can work well
  • Weak method (makes little use of knowledge)

State Spaces

  • Initial state and goal state, the path along is a means of solving the problem

Means-Ends Analysis

  • Positive steps reduce the difference between the current state and the goal state, think the a* algorithm
  • Example of a universal AI method
  • No guarantee of optimality, or even being able to solve the problem

Problem Reduction

  • Decompose a hard problem into several smaller, easier problems

  • Think about how recurrsive solutions work to solve problems

  • In the block stacking example, decompose the final state into a series of "milestones" that are more likely to lead to a successful solution.

  • Does not provide guarantees of success