Week 8: Journal Questions
Answer the following question for
Monday's reading on Genetic Algorithms (Mitchell GA ch.1, PDF in course Google Drive folder)
-
In your own words, explain the appeal of applying evolutionary computation to computational problems.
-
Genetic algorithms are another form of search algorithm. Of all of the search algorithms we saw in the first half of the semester, which one is most similar to GAs and why?
Answer the following question for
Wednesday's Evolving Neural Networks (Mitchell GA section 2.3, PDF in course Google Drive folder)
-
One way to evolve neural networks is to pick a fixed network archietecture (the number of input units, hidden units, and output units), and to use a GA to discover a good set of weights.
- How are the chromosomes represented in this scenario?
- What is a typical fitness function used in this scenario?
-
For what kinds of tasks would we prefer evolving weights over using backpropagation to learn weights?
For Friday's reading there's no required response; do your best to read the paper, but don't be afraid to skim over some of the math if you find it confusing, and don't worry if you don't understand every technique that's referenced.