The Objective Function for this application is all the contraints used
to match students and dorm rooms. The domain is all the students in
beds. A single data point, or *state*, is the assignment of each
student to a bed. maps from a state to a non-negative integer
score, where a zero score (if you could get there) would be `perfect'
with no errors, and everything else is worse. Starting at some state,
we try to improve it by moving to nearby states. By switching a
randomly chosen student with the occupant of a randomly chosen bed,
while all else stays constant, we can reach a nearby state. Each
state has some four million possible neighbors, since any of almost
2,000 students can be moved to any of 2,000 beds. To better illustrate
the concepts of *neighborhood* and *landscape*, we present
graphs showing a single data point from both random and optimized
configurations.

elena s ackley 2002-01-20 download thesis