Name: ___________________

Student No: ___________________

Faculty: ___________________

Signature: ___________________

THE UNIVERSITY OF QUEENSLAND

FINAL EXAMINATION, SEMESTER II, NOVEMBER 2002

COGS2010

LABORATORY INTRODUCTION TO MODELS IN COGNITIVE SCIENCE

Time: Two (2) hours for working

Ten (10) minutes for perusal before examination begins

General Directions to Candidates

There are twelve questions on the paper. You should attempt all questions. Each one is worth 10 marks (Total marks 120).

The examination has been set to take the full two hours, so answer the questions that you know first, and only after answering them, go back to the questions that you are less certain about.

Answer all questions on the examination paper. If you need additional space use the back of the paper or an additional exam booklet and clearly label each answer.

Question 1

marks

a) Consider an interactive activation and competition (IAC) network that represents clothing outfits. The network has pools of features for {tshirt, business shirt}, {shorts, jeans}, {thongs, running shoes}, {socks, no socks} and people’s names. Draw an IAC network that will represent the patterns for (1) Alice’s clothing (tshirt, shorts, running shoes and socks) (2) Bob’s clothing (tshirt, jeans and thongs).

6

b) How many positive and negative weights does your network have? Explain your answer.

1

c) If another 50 clothing patterns were added to your IAC network, it would be possible to use the network to generate prototypes. Explain how you would use the network to generate the typical shoes worn with jeans.

3

Question 2

marks

a) Describe the Stroop effect.

4

b) Explain the role learning plays in Cohen, Dunbar and McClelland’s theory of the Stroop effect, and how it was implemented in the model.

3

c) Explain what training and test patterns were used in the model, and why they were used.

3

Question 3

marks

a) Explain the difference between distributed and local representations in the input and hidden layers of a neural network.

2

b) List one advantage of each for input patterns.

2

c) Give an example of an error correcting learning rule for a one layer neural network.

2

d) Consider a feedforward network that has one input, one hidden and one output unit (a 1-1-1 network). The training patterns are the identity patterns (0,0) and (1,1).

i How many local minima does this network have?

ii How many of these are global minima?

2

e) From a memory modeling perspective, what problems would prevent adding a hidden layer to a matrix memory model?

2

Question 4

marks

a) Explain what the XOR problem is, and why it is relevant to machine learning.

3

b) Can a one-layer feedforward neural network learn the XOR problem? Justify your answer

2

c) In general, is backpropagation guaranteed to converge to

i. A local minima

ii. A global minima

Justify your answers.

3

d) Would the Hebbian learning rule be appropriate for training the 1-1-1 network? Justify your answer.

2

Question 5

marks

a) Consider four patterns, (0,0), (1,0), (0,1) and (1,1). If a 3x3 self-organizing map (SOM) were trained on these patterns, what would a likely SOM layer look like? (Draw the input and map layers)

5

b) What would the trained SOM do with the patterns:

i. (0.5, 0.5) Justify your answer

ii. (1.5, 1.5) Justify your answer

3

c) Explain what the neighbourhood parameter does in a SOM.

2

Question 6

Marks

a) Explain what an analogy is in cognitive science.

2

b) The structure mapping engine (SME) is an analogical mapper that takes hand-coded representations of source and target domains and finds mappings based on the similarities between the given representations. How does Copycat differ from this approach when making an analogy with respect to its mapping and representation formation?

2

c) Is the parallel terraced scan in Copycat a strategy that searches breadth first, depth first or neither?

1

Copycat question continued:

A possible solution derived by Copycat to the problem abc : abd, ijjkkk : ? is ijjlll. This solution is driven by the fact that both abc and ijjkkk are successor groups to the right, allowing a correspondence to be formed between them. For the solution ijjlll to be formed in the Workspace:

d) what initial and modified rule would allow this to occur ?

1

initial rule: replace ________________________ by _______________________

modified rule: replace ________________________ by ________________________

e) The slippages between the concepts used in the initial and modified rule are either implied by or incorporated in the correspondence that is built between the groups abc and ijjkkk. Denote what descriptions would be attached to the two groups, the slippages that occur when the two groups are mapped, and what implied slippages are required to derive the modified from the initial rule above.

1

Copycat question continued:

f) In the following diagram, fill in the missing descriptions, groups, bonds and correspondences that are consistent with your answers above.

3