1.
All population-based
evolutionary algorithms have an exploration/exploitation dilemma: How much
should the population explore widely across the space, vs
how much should it concentrate its population (and hence search) in the
neighbourhood of the best solution found so far.
a.
What
examples from nature might favour broad diversity?
b.
What
examples might favour specialization?
2.
Some EC techniques use
only mutation as their source of variation.
For these techniques, populations of a single individual are often used,
with multiple restarts.
a.
Could you add
crossover to such a technique? Why or
why not?
b.
What
strategy does hill-climbing with multiple restarts employ in exploring vs exploiting a fitness space?
3.
Selection
techniques:
a.
Describe
roulette-wheel selection.
b.
How does rank
selection differ?
c.
What
strategies do they employ in the exploitation/exploration tradeoff?
1.
What is a biomorph? Why is it of interest to cognitive science?