Homework 3

Learning

810:161

Artificial Intelligence

Fall Semester 2001

Due: Friday, November 30, 4:00 PM

1. Consider the problem faced by a person learning to play tennis (or some other competitive sport with which you are familiar). Explain how this process fits into the general learning architecture that we discussed in Session 15. Describe how each of the components of the architecture operate when learning to play. Is the learning primarily induction learning or reinforcement learning?

2. Use the induction algorithm discussed in class to induce a decision tree from the data given in this table:

      Apples  Beef  Cake  Dates  Eggs  Fish  Grapes  Honey  |  Reaction
      -----------------------------------------------------------------
      yes     no    yes   yes    yes   yes   yes     yes    |  yes
      no      no    yes   yes    yes   yes   no      no     |  no
      yes     no    yes   yes    yes   no    yes     no     |  yes
      yes     no    yes   no     no    no    no      no     |  no
      yes     no    yes   yes    yes   yes   no      yes    |  yes
      yes     yes   no    yes    yes   yes   no      no     |  no
      no      yes   no    yes    yes   yes   no      no     |  no
   

3. In our class examples, we never tested the same attribute twice in one branch of a decision tree. Why not? Under what circumstances might we want to test the same attribute twice in one branch?

4. You could use the simple form of genetic programming we studied in Session 19 and Session 20 to learn a function that matches the data in the Problem 2.

   • Give a representation for the "programs" that the system will manipulate.

   • Give two different fitness functions for the programs. Which do you think is better? Why?

Please submit a hard copy of your solution only--no e-mail this time.
You may submit your solution to me in person or to the department secretary (219 Wright).