Questions:
For the second test: note
that there are 22 questions; you can expect to be asked to answer 6 or 7 out of
8 or 9.
- List
three distinct problems Clark mentions with Marr’s three
levels. Explain briefly.
- What
does it mean to say that the world is its own best model? (see Clark, page 91).
- Describe
and explain a case where Clark thinks one needs representations and one
where he thinks one does not.
- Why
doesn’t Clark think the brain is best characterized as the seat of
reason?
- Describe
why Clark does not think methodological functionalism should be the
premiere approach in cognitive science.
- Explain
the difference between weak and strong equivalence. Why doesn’t strong
equivalence require that equivalent systems are made of the same material?
- What
is methodological functionalism?
- How is
folk psychology connected by Dawson with the computational level?
- How
did Hume account for the fact that actions seem to get explained in terms
of future possibilities?
- Why
does Dawson think reverse functionalism is essential for connectionists?
- Consider
a neuron obeying McCulloch-Pitts equations. Show that this neuron can only solve classification
problems that are "linearly separable".
- Define
classical conditioning and operant conditioning. Emphasize their difference.
- Which
components of a neural network are generally used for long-term memory?
How does learning occur?
- State
"Hebb's postulate" and discuss the conceptual framework
underlying this postulate.
- What are the limitations of currently available neural
networks in learning arbitrarily complex problems?
- Illustrate
the difference between heuristic and algorithmic reasoning by way of an
example, but please use an example different from the one(s) found in the
notes.
- A
group of graduate students has developed a knowledge based-system for
class scheduling. All one has to do is to supply the system with the
classes to be offered in a given semester, the available class rooms, and
the professors and their specific expertise (plus teaching load). The system
will come up with a class schedule showing who will teach what, when and
where.
- Discuss
the knowledge base needed to solve the class scheduling problem. Provide
examples of the facts and rules one would expect in such a system.
- Suppose
you want to extend the system in such a way that it will not schedule two
classes at the same time, if these classes are likely to be taken by
students in the same semester. What knowledge would you add?
- Discuss
whether the class scheduling problem (i.e., figuring out who should teach
which class at what time in which room) is a forward chaining or backward
chaining problem (or a combination of both).
- Describe
the components of a knowledge-based system and discuss how realistic it is
to use such systems as a metaphor of human intelligence.
- Describe
the design of an intelligent system to determine if a particular artist
produced a painting. You may assume that the artist has produced a fairly
large number of paintings. Discuss such issues as the knowledge
representation and reasoning strategy you would use.
- Provide
an example of a semantic net. Please use original examples, i.e., not ones
found in the notes or discusses in class.
- What
learning strategy is (most likely) followed when acquiring the following
skills:
a. Learning to drive a bicycle
b. Learning a foreign language
c. Learning the way around town
d. Learning to speak