Re: Robustness Check and "A growing body of ad-hoc analysis solutions"

[Benedikt Stefansson]

Marcus G. Daniels wrote:
> 
> >>>>> "B" == Benedikt Stefansson <address@hidden> writes:
> 
> B> Sometimes it is just easiest to go the hardcode way
> 
> Don't go the hardcode way unless you're comfortable scrutinizing code.
> If you don't read your code and treat with the kind of respect you
> give your papers, you will have code worthy only of disposal.
> 
> There's nothing wrong with throwaway code, of course, but decide what
> mode of behavior you want given your circumstances.  What you *don't*
> want to do is have inadequate mental energy for maintaining a growing
> body of ad-hoc analysis solutions within your simulations, and have
> them give you flakey output. 

To clarify my statement, you can save time on retooling and analyzing
outside of Swarm by coding in the calculation of summary statistics, but
don't write simulations which don't allow you to switch back to full
output mode.

If you save the random seeds for each run and have a switch in the
program that either spits out full output or the summary statistics you
can always move between "debug" mode and "get that paper done" mode. 

> Also, don't trust prototype software like ExperimentSwarm.  Don't even
> think about using such code unless you are prepared to start reading,
> contemplating, and modifying that code.

Any beta/prototype code that goes from end user to end user should be
treated with caution until it has been used for T time periods by N
drones. But this is after all what Swarm should be about, ne?

For someone that has been cursed by trying to teach Swarm (whatever that
means) it is a bit embarassing that we don't provide some very basic
DroneLite capability in Swarm (read setup file for parameter sweeps from
file, control model runs iterating over parameters). 

All this ExperimentSwarm thing which I alluded to is supposed to do is
read in a file like

@begin
a x_a y_a z_a
b x_b y_b z_b
c x_c y_c z_c
@end

where a,b,c are parameters of Model, x_i is start value, y_i is stop
value and z_i is step value for iteration of parameter i in {a,b,c}. It
then spits out the model.setup file, creates and primes model, runs
model, drops model, rinse and repeat.

I would be happy and eager to enter into discussion about the
standardization part (i.e. "growing body of ad-hoc analysis solutions"). 

For example, is the '@address@hidden' approach dead? Which simple ASCII
format could become the chosen serialization file format and an eventual
parameter sweep file format .

Keep in mind the lowest common denominator set of knowledge of other
tools/languages. Great to leverage off Scheme/HDF5/R but pedagogically
speaking they make beginners want to run for the hills.

It just so happens that Swarm is now used on the little known and
esoteric "Windows" platform (there may be some information about it on
the web for those who are unhappy with their Unix or Mac workstations).
This Windoes thing is popular with users which don't have numerous
programming languages under their belt, and expect this Swarm thing to
provide rudimentary capabilities out of the box. I think it makes sense
to provide Drone like capability in what is being presented as a basic
toolbox for simulation.

Regards,
Benedikt
 
-----
Present coordinates: 
Dep. of Economics, Univ. of Trento, Via Inama 1, 38100 Trento, ITALY
Off: +39 0461 882246/267875 Mob: +39 347 0415721 Fax: +39 0461 882222

                  ==================================
   Swarm-Support is for discussion of the technical details of the day
   to day usage of Swarm.  For list administration needs (esp.
   [un]subscribing), please send a message to <address@hidden>
   with "help" in the body of the message.