Announcement
What: Resume Critiquing Workshop
Bring: A current copy of your resume
When: 4:10 p.m. - 5:00 p.m. on Friday, January 27th
Where: 350 EPS (across from the CS Conference Room)
Special Guest: Doug Warner of RightNow Technologies
Sponsor: Local ACM Chapter
Chapter 3: GAs in Scientific Models
Section 3.2: Modeling Sexual Selection
- Collins and Jefferson (1992)
- The male has a trait, t, that is either absent or present.
Unfortunately, trait t is maladaptive and increases the likelihood
of the male's early demise!
- The female has preference, p, for the male trait above
- Population size = 131072 (217)
- Mutation rate = 0.00001 / gene
- Number of generations = 500
- Each female crosses probabilistically with one of a small number
of surviving males
- Figure 3.10 plots t = 1 vs. p = 1
- General observation: it is tough to balance simplicity
(to make the model understandable) with generality
(to make the results of the model meaningful)
Section 3.3: Modeling Ecosystems
- Holland first created the
Echo
system in 1975
- Agents can mate, trade or engage in combat
- Figure 3.11 shows the abundance of different genotypes.
It is similar to some real world ecology plots.
Section 3.4: Measuring Evolutionary Activity
- Bedau, Packard (1992)
- The idea is to measure evolutionary activity by looking
for evidence of persistent usage of new genes
- The Strategic Bugs World consists of bugs and food on
a 2-D lattice
- Each bug takes inputs from five surrounding squares. Each input
consists of two bits: least food, less food, more food, most food.
Thus, there are 210 possible inputs.
- Each bug has two outputs: the direction to move (four bits) and
the distance to move (four bits)
- Figure 3.12 shows the chromosome representation
- The population size is 50. Each bug is a partial, randomly
assigned lookup table. (A complete table would consist of
210 * 8 bits yielding a search space of size
2213.)
- Each gene has a counter associated with it. The counter is initialized
to 0 and is incremented each time the gene is used.
- Figure 3.13 shows usage statistics for one run. White indicates
no usage, black indicates maximum usage.
- Define μ0 to be the baseline usage of a gene.
In other words, it is the usage a gene would obtain if selection
was random instead of fitness based.
- Define net persistence, P(t, μ), to be the proportion
of genes in the population at time t that have usage μ or greater.
- Define evolutionary activity, A(t) to be - [ δ P(t, μ) /
δ μ ] when μ = μ 0
- Claim: If A(t) > 0, then evolution is occurring
- Claim: If A(t) > 0, then the system exhibits life!
