Final Exam Review Questions, CS 2400 (Hasker)

The exam will be open-book, open-note, but individual. Do not discuss on exam content with anyone until the exam is graded.

See also the midterm review questions

Review questions (for reference only - the exam may include other topics!):

1. How does a neural network mimic the brain?
2. Give one way a neural network differs from the brain in structure.
3. What is the impact of not allowing hidden layers in an artificial neural network?
4. What is the sigmoid function and why is it an interesting activation function for neural networks?
5. Know how to compute logic functions using perceptrons.
6. Describe feed-forward and back-propagation in neural networks.
7. Why do we use derivatives during back-propagation?
8. Why is the chain rule often necessary for back-propagation?
9. Describe stochastic gradient descent in a sentence.
10. Give a sentence-long description of reinforcement learning.
11. When should reinforcement learning stop?
12. How does the value function differ from the reward function in reinforcement learning?
13. Is the computer harmed by reinforcement learning?
14. Why is a discount factor important to reinforcement learning?
15. What would be a case where mutation is critical to genetic algorithms?
16. Describe the difference between genetic algorithms and genetic programming.
17. Why might it be unwise to allow mutation to be applied to 25% of the population in a genetic algorithm?
18. Explain what it means for a learning algorithm to be "robust to noise."