Framsticks

Hi.
I'm studying NN... looking for an interactive place to try my researches,
i've found Frasmesticks. I looks really nice. I can create "static"
creatures well using the f1 genotype. I am writing to ask some help about
NN... i've read all the examples and explanations, but i can't understand a
lot of things... for example:

How can I train the Net? or am I supposed to write the correct weights for
every synapse for the job?

Where is a good newbie tutorial for NNing in framsticks?

thanks!

I'm new at this so your patience please.
The gene pool population keeps growing and growing even though the gene pool
capacity is set.
The capacity may be set at 200 but the population will quickly surpass this
number, eventually being many thousands.
The easiest example I can give is to simply start F2.2 and run, select f1
for seed, and wait. The gene pool cap is set at 200 but this will,
seemingly, be quickly ignored.
???

I have asked about this before, but since trying it again on a win98se
machine I thought I'd ask again.

I have noticed that movement in a circle is measured as distance. How about
a new measure; distance from the point of origin?

Also, has anyone noticed that the ground level is a little high (the sticks
seem to sink into the ground).

Hi,

2.2 is looking great!!!

But I keep getting "Invalid Floating Point Operation" , then I have to close
Framsticks and start again to carry on.

There seems to be no real pattern that I can see.

Any thoughts?

Ander

Hi All,

Since my experience with Framsticks has just begun this could very well be
something I'm doing wrong but I thought I'd ask anyway...

If gene pool capacity =200 then why do the genotypes in the genepool
window keep showing increasing numbers of genotypes, 12000+ for example?

Program grows increasingly sluggish sometimes. Example: simulating 3
creatures at a time on flat land with selection for say vertical position
and pool capacity at 200, after ~25M steps there are 12K+ genotypes in the
gene pool window, the program still runs okay, but if changes are made to
selection, or deletion of some genotypes, or clearing of performance then
the program is likely to become *very* slow to respond to nonresponding.
This is on WinXPhome, P3-1GHz, 512MB. Memory is not the problem, there's
still plenty left open when this happens. Unregistered (for now) version
2.2.

If you decide to tweak the fitness parameters during an experiment, how do
you properly clear out the old fitness values for the current genotypes?

Thanks!!

David
Rochester NY usa

I am trying to unravel the meaning of different collision masks.
For example: The standard experiment definition states for Creaturegroup1,
frams, a colmask of 13 (the sum of 1+4+8) and for Creaturegroup 2, food, a
colmask of 148 (= 4+16+128).

How can I find out the different functions of the individual colmasks
(1,4,8,16, ...,128) without having to test each combination? Is there a
logic to it that I missed?
Or could somebody provide a list of the functions of the colmasks?

Thanks in advance,

Walter

This is probably an extraordinarily stupid question, but: why don't I
see any energy balls in my world display? I've spent a lot of time
hunting for a Parameters setting to spawn more food, but all I've found
is an apparently unmodifiable "Food" population, and the setting for
energy per (nonexistent) ball.

TIA,
- Sam

Hi there,

I did a lifespan directed evolution experiment on a genepool containing
food finders and fast walkers. I kept a constant number of 10 energy balls
in the world. All went well for a 100 million steps or so, and then the
inevitable happened. One instance of a mutant proved so effective that it
could gather food more rapidly than burn it. The simulation ran unattended,
and since energy balls are replaced immediately after consumption, the
bastard had aggregated over a million of units of energy, taken in at 200
units a ball, and I had wasted a night or so of computations on it.

Question is, how can I avoid this? Can I set a maximum value for lifespan,
some kill_at value, and possibly credit the fram's fitness rating for the
remaining amount of energy?

Lowering the energy content of the balls is not an option. This would work
too much against the other creatures. Also, it is no guarantee that this
will not happen again. This was instance number 233 of this same genotype.
It's 232 brothers or sisters (are frams male of female?) had an average life
span of 13500 steps, that is just 35% over the basic lifespan in the
experiment.

It is obvious that this instance just had a lucky draw. It would be helpful
to have individual statistics foe each genotype, giving minima, maxima and
standard deviation of its instances. I assume this will be easy done with
some script, once version 2.2 is out......

Frans Verbaas

Jerome Fulachier Wrote:

Hi all,

if you want, you can look at my online experiments results page:

http://isnpx0168.in2p3.fr/F/

Give me feed back if you are interested!

Is it possible to output more detailed periodic 'backups' of the experiment
state? Values like the fitness of each genotype in the pool?

Also, what exactly do the names of the Frams signify? What do Frams with the
same last name have in common?

I assume it vaguely corresponds to a 'species' of fram, although I am
assuming it is possible to get fairly distinct frams within a 'species' as
given by the name? Or , rather than species, does it tell me more about a
'hereditary group'? (Please excuse my terminology, I hope the concept is
getting across..)

How does the meaning of the second name change if similarity is set at say,
0.3?

Thanks,

Matthew.

---
Outgoing mail is certified Virus Free.
Checked by AVG anti-virus system (http://www.grisoft.com).
Version: 6.0.391 / Virus Database: 222 - Release Date: 19/09/2002