|
How
evolutionary ecological simulators model the world for evolution@home
The
world is hopelessly complex. Human minds have no chance to understand
it all at once. Thatās why scientists build models. Each model
is a simplification of reality and tries to catch the essence
of the most important factors while ignoring all the details
that are believed to be unimportant.
This means that the same biological reality can be described
by many different models. While you can test the quality of
models, you can never prove them. Ultimatively, however, those
models will be successful that are most difficult to disprove
by empirical data. To find out how reality is best described,
relevant models need to be tested.
And that is a complex task. This is why each model, describing
an evolving world, gets its own simulator to facilitate computation
of the models predictions. In short:
- 1 simulator x =
- 1 biological model x =
- 1 executable programm x
running
- many single runs =
- many simulations =
- many parameter combinations =
- many "elementary scientific questions of type x"=
- many simulate-commands framed by & and # in
a run-file
Each such model is comprised of several input-parameters
that govern the fate of the model world while the interesting
output-parameters are observed. A model with even a few input-parameters
demands a huge number of simulations to get an idea of the
structure of multidimensional parameterspace for that model.
This task is more than any single computer can handle. Therefore
the task is distributed to many computers. Technically speaking,
the simulators use individual-based models for their stochastic
simulations.
Since many different models are possible, different simulators
will be published on these pages and you can chose into what
model you want to invest your CPU-time in.
How
to run a simulation manually
First of all,
just select a simulator and download it. Put it into a special
folder (like evolution@home on your desktop) and enter your
preferences after the first start: You will have to tell the
simulator an upper limit for the RAM-size and computing time
you are willing to commit to a single simulation (not the computing
time of a whole run-file). Do not use virtual memory
by trying to compute simulations that need more RAM than
actually available on your system. Virtual memory will slow
down your sytem and render it virtually useless. After setting
these upper limits, you can personalize the results you will
compute by entering group and personal names, a web-link you
want to promote and PINs that prevent other people to submit
results (and links) under your name.
Then you are ready to download a run-file of the appropriate
RAM and time complexity. Save it in the simulators folder
under the name "run" or "run.txt". Now
you are ready to start the simulator. If you want to optimize
your computing performance, switch off screen savers and energy
saving options. Furthermore put a link (alias) to the autostart
(startup) folder of your operating system: this will automatically
start the simulation after booting.
If you want to submit results (suggested when the
results file becomes large (e.g. 1MB) or more than 2 weeks
have passed), remove the results-file (you will recognize
it by its name) from the simulators folder and email it as
an attachment to the address specified for that simulator.
Do not remove results twice and do not move the results-file
back to the simulators folder as this would result in resubmission
of these results.
Repeats are not duplicates: As results will be checked
for duplicates, it is useless to submit the same results-file
several times. However, different runs of the same simulate-commands
are not duplicates, as these get different random seeds and
thus lead to different results. Such repetitions are important
to calculate elementary statistics like mean and variance
for the parameters observed. Once a parameter combination
has been repeated often enough, it is removed from the run-files
on the website.
Once all simulations in a run-file have been completed,
the simulator quits automatically and renames the run-file
to completed-run-file. Then return to this website to get
the next run-files. You may also compose your own run-files
by combining various simulate-commands from the published
run-files into a single run-file by copy and paste.
While full automated distribution of keys and submission
of results is not possible yet, this manual method gives you
maximal flexibility in choosing the type of simulations you
want to run. For more details consult the documentation included
with the simulator.
|
