Describe the structure of a semantic network.
Explain how individual ideas or concepts are represented in a semantic network.
Describe the concept of "activation levels of nodes" and explain the effect of the level of activation.
Describe what determines the level of activation of a node.
Explain how knowledge about concepts is represented in a semantic network.
Explain what determines the strength of an association between nodes in a semantic network.
Describe what determines whether a node sends activation to other nodes.
Explain the concept of spreading activation.
Explain how hints, context, and mnemonics improve memory by providing more connections or acting as "parasites" on nodes with stronger connections.
Describe a lexical decision task.
Explain how a lexical decision task is used to test predictions about spreading activation.
Describe a sentence verification task.
Explain how sentence verification tasks are used to test predictions about the organization of knowledge in a semantic network.
Define "degree of fan".
Explain the implication of a high degree of fan for the spread of activation.
Name three different ways that a search through a semantic network memory system could begin.
Describe a "winner takes all" system for explaining how a search through memory leads to activation of a particular target node.
Define a "proposition".
Describe the types of links in Collins and Quillian's, Collins and Loftus', and Anderson's ACT theory.
Describe the kinds of nodes in each of the above.
State a problem that is not dealt with very well by semantic networks such as Collins and Quillian's, Collins and Loftus', and Anderson's ACT theory.
Define a "connectionist" or "PDP" model.
State the term used for the strength of the connection between two nodes in a connectionist model.
Describe how a concept is represented in a connectionist model.
(Examples of concepts:
The name of your dog, Skippy, or Rex
The characteristics of Skippy, or Rex
The appearance of a rose, or a steak
The smell of a rose, or a steak).
Describe how knowledge about concepts is represented in a connectionist
model.
(Examples of knowledge about concepts:
Your knowledge that Skippy is one kind of dog
Your knowledge that Rex is a different kind of dog
Your knowledge of what a rose smells like
Your knowledge of what a steak smells like).
(Click here to see an example of a connectionist network illustrating your knowledge of the characteristics of your two dogs)
(Click here to see an example of a connectionist network illustrating your knowledge of what a rose and a steak smell like)
(Click here to see a working model of the network illustrating knowledge of what a rose and a steak smell like)
Explain what an "emergent property" is.
Explain the emergent properties, "content addressability" and "spontaneous generalization".
(Click here to see an example of a connectionist network that illustrates content addressability and spontaneous generalization).
Describe the process by which a connectionist network "learns".
Describe the emergent property, "graceful degradation".
(Click here to see an example of a connectionist network that illustrates learning and graceful degradation. It uses the "rose and steak" example that you saw above, but it is a computer simulation, not just a diagram. Run the simulation.)
How would a connectionist model solve the XOR problem?
List some applications of neural networks.