
Parameter sampling evolutionary experiment.
Goal: to gain knowledge about the influence of parameters
on evolutionary optimization processes.
Optimization course at CS/PUT, supervised by Maciej Komosinski.
---------------------------------------------------------------

Students work in teams (2 or 3 people in each).
Each team uses a different genetic representation in their experiments.
2-student teams: genetic formats 2,3,7
3-student teams: genetic formats 0,1,4


Each student tests:
- a single selection/deletion setting
- a range of stagnation periods (from 20 to 5000 step 20, processing logs obtained for stagnation=5000)
- four population sizes
- and repeats the experiment at least 3x, at most 10x for each population size
  (thus performing a total of at least 12, at most 40 runs).
- note: for some parameters, experiments may take a long time.
  Start early!
  
  
Each student prepares a written report with
- a 3D chart: average fitness vs. popsize & stagnation
- a 2D chart: for longest stagnation, minimal/average/best fitness vs. popsize
- a 3D chart: average total number of evaluated creatures vs. popsize & stagnation
- a few charts: fitness vs. time
- a few most interesting constructs
- *conclusions*
Time to prepare: 2 weeks.


Each team prepares *working together* a summary of all
experiments and a presentation (~15 min) with
- charts summarizing all experiments for all desirable parameter combinations
- final conclusions (comparison of results and relation to optimization parameters)
- most interesting constructs.
Say a few words about what software you used to process, handle and visualize large
amounts of data. Highlight two values: for longest stagnation and biggest population
size: average best fitness and average number of evaluated creatures.
Time to prepare: 1 week.



******************************** UPDATE: ********************************

1. Info about genetic encodings: (say a few words during your the presentation)
   f0,f1,f4: http://www.frams.alife.pl/a/al_genotype.html
   f2,f3: http://www.frams.alife.pl/wiki/index.php/Biologial_representation
   f7: http://www.frams.alife.pl/wiki/index.php/MessyGenetics

2. Making 3D screenshots of the most interesting genotypes:
   Use Framsticks Viewer from http://www.frams.alife.pl/p/al_download.html
   It only supports f0, f1 and f4 BUT for f2/3/6 you can get the
   converted f0 genotype (in Framsticks GUI) and feed it to framsview as usual.
   OR, 
   if you like to/can hack and want nicer pictures, get the Framsticks
   v3.x server from http://www.frams.alife.pl/dev/files/
   send the genotypes to the server (e.g. using telnet), revive them 
   (see samples at http://www.frams.alife.pl/common/server.html)
   and then use
   http://www.frams.alife.pl/wiki/index.php/Client3D
