Final exam information
- The exam is Wednesday December 15 at 2pm in Amos Eaton 214.
- The exam is closed book and closed notes. (However, see "tips" below.)
- You may want a calculator for some basic probability
calculations, but it won't be anything you can't handle by hand.
- Topics from sections of the text we have covered, topics covered
in class, topics covered in handouts, and topics covered in
assignments are all fair game.
- The exam is comprehensive, though it focuses more on the second
half of the class.
Tips, hints, and other information???
I have written 8 problems for the final exam. However, I am still
adjusting the length and difficulty of the exam, so bear in mind that
my comments are about an exam that is not finalized yet. Therefore,
the general comments are more likely to be true than the specific
comments. I'm not inclined to reveal too much about the exam, but I
will tell you the following:
- There aren't many "apply this algorithm to this input" or simple
"recall" questions on the exam. Instead, I favor questions that
assume you know the algorithm or concept and ask you do some analysis
or evaluation to demonstrate your understanding.
- You should understand why and how the learning algorithms we covered work.
- There is a constraint satisfaction problem.
- I will provide the "long" equations from learning algorithms.
These include the TD Q-learning update rule, the Neural network
backpropagation equations, and the perceptron learning rule. My focus
is on understanding concepts and algorithms rather than memorization.