Homework 3 - Chapters 5 and 6

Due: Friday February 8, 2008 at 5pm

Note: There was a typo on Question 7 originally. It has been corrected below.

Each question is worth 2 points.

  1. 5.1: What is a frame? What are the class and instance? Give examples.
  2. 5.8: What is a method? What are the most popular types of methods used in frame-based expert systems?
  3. Describe how JESS slotted facts differ from the frame-based expert system described in the book.
  4. Define the three main types of learning: supervised, unsupervised and reinforcement.
  5. How does an artificial neural network model the brain?
  6. Why can the perceptron only learn linearly seperable functions? Give an example of a linearly seperable problem and a non-linearly seperable problem.
  7. Design a two-input perceptron that implements A XOR OR B.
  8. Define feedforward and backpropagation. How do the methods differ?
  9. What is a multilayer perceptron? Be sure to explain what the hidden layer is for and what it hides.
  10. 6.5: What are the main problems with the backpropagation learning algorithm? How can learning be accelerated in multilayer neural networks? Define the generalised delta rule.