1. State Spaces
(a) A chess board has 64 squares and there are 32 different pieces. Assuming that each piece can be in any square, how many different states does the game have?
(8 marks)
(b) Describe, using code or pseudo-code, an algorithm to search for a solution to a travelling salesman problem with the shortest path.
(9 marks)
(c) There is one algorithm that is particularly good for playing adversarial games like chess? Name and briefly explain this algorithm. What is an essential component for this algorithm? Name one typical mechanism used to improve this algorithm.
(8 marks)
2. Knowledge Representation
(a) Write a semantic net for the topics in CST3170. This should include at least 12 nodes, and 5 types of arcs, including the most important types of arcs. In this case, the topics are also important.
(10 marks)
(b) Is XML Turing Complete? Another way to phrase this is can you compute everything in XML that you can compute in Java? Why or why not?
(8 marks)
(c) How do rules in a rule based system differ from predicates in first order predicate logic?
(7 marks)
3. Machine Learning
(a) What is the prime difference between multi-layer perceptrons and deep nets? How can this difference lead to an improved performance on big data tasks?
(8 marks)
(b) Explain how new members of a population are created in genetic algorithms?
(9 marks)
(c) What is a kernel function, and when is it useful?
(8 marks)
4. Applications
(a) Parsing is a major component of syntactic processing, is one commonly used sub-system in natural language processing applications. Name and briefly describe three other common sub-systems.
(9 marks)
(b) Name four sensors and four effectors used by robots. (They don’t have to all be used by the same robot.)
(8 marks)
(c) Shallow natural language processing systems often use bag of word techniques to approximate semantics (the meaning of words and texts). Describe how these techniques might work.
(8 marks)