Thursday, February 27, 2003

at poly atm, waiting if david cud help with registering my ethernet card today.

here's a picture of voronoi method of spatial analysis in action.
taken from p.73 a+t 19 Spring 2002

                    urban application of voronoi method    the magazine

its a very nice application of voronoi method to real spatial problem, in this case, urban scale. they (people who develop KaisersRot software) quite confident about it for enabling the generation of urban plan which is based wholly on community demand p.74.

about voronoi method, here's different sort of applications
http://www.ics.uci.edu/~eppstein/gina/voronoi.html

so p asked me to start by choosing which sort of geometry it will be for the project. this comes down to two choices, orthogonal or free form.
voronoi more likely be the method to go on if i choose free form geometry. if orthogonal, perhaps shape grammar is sufficient.
having listen to p and read abit more, i think free form by voronoi is more interesting to try.

and about the structural relationship between objects in simulation, again, i sense that p wants to push forward to coevolutionary model. which i agree, of course.
but my personal opinion, is at the moment, it will be suitable to try to evolve spatial configuration based on human animat, and let human animat not-evolving. until i managed to develop this, i think i dont mind to go on with coevolutionary model.

some metaphysical assumption also has to be decided to somewhat direct the evolution to 1. evolving into a rave, large one cell type of form, or 2. into a kind of monastery, ie. one cell per each agent. on which i think spatial configuration is something in between. the victorian plum tree story is interesting because it can be used to direct evolution into something we suspect it will be. if more fruit and less branching is the sort of fitness that gardeners looking for, perhaps that we could also find in what range spatial configuration is; in between those extremes.

at some point this week i have to be able to decide human animat modus operandii, in which should be very simple.

days to consider for regular meeting:
monday & tuesday, after lunch because morning with tara.
friday afternoon dont clash with anything.

Wednesday, February 26, 2003

uploaded presentation date 24 feb 03
go to me::press (yup!! click the frog please)
mind u with 56kbps modem, it'll take quite alot of time to upload coz i put fancy schmancy ppt animation in.
perhaps i shud zipped it. if anybody ask..

Monday, February 24, 2003

tomorrow poly ...
missed!!
thursday then

Thursday, February 20, 2003

p's verdict on my page SCARY *LOL* what's scary; color coordination or those notes i put down in here?

from p 11.24
i had a look at your weblog last night - its really scary!
I have decided that all this talk of classifier systems is a complete red
herring - all evolutionary algorithms can be classifier systems if they want
cant they? anyway its about time you started doing some real work and then
we can work out whats what.


i agree with lets do some work (this means, the programming bits... it hasnt started ...*sigh*)
perhaps is a red herring, but anyway it gives me clear idea how to use EA to do something.
me personal opinion, it answers encoding a problem, then using it as string and do evolutionary procedure with it.
and its because based on classic if-then rules, CS release us from problem of decoding the string/what to do after reproduction.
i agree all EA can be CS, simply because CS is how we use EA.

p, join the blog!! your comments'r surely welcome

Wednesday, February 19, 2003

ftp'd
Classifier System in Swarm

http://cscs.umich.edu/Software/CFS-S.html



where's the documentation then?

i emailed rick riolo 17:23, and his reply:

There was no particular motivation or associated
paper for that project...we were just trying to
see how it would be to implement a general-purpose
classifier system in swarm/objective-C.
to learn about classifier systems in general, visit:
http://www.cs.bath.ac.uk/~amb/LCSWEB/
I think there may be some other classifier systems
in Swarm now, as well, perhaps being actively used.
See the Swarm.org user community and contribution pages.
- r



inventors and engineers.... inventors just invent, then the engineers who make use of inventions.
well...

ok r, i found a lot just have to pick the rite one, or not...
http://www.swarm.org/community-contrib.html

Tuesday, February 18, 2003

the thread...

configurational theory of architecture >> autopoiesis >> classifier system


1. circularity
   the nature of explaining phenomena with autopoiesis framework. why? because all in all, it depends on the observer.
   configuration of space influence and can be influenced by configuration of people.
   cs and its environment. cs acts and reacts to environmental inputs.
2. the frog
   animat in cs example page
   maturana example of frog's retina
   frog's configuration? eh eyyyy...

i need several chapters from autocad 2002 complete by Sybex 2002 copied. anybody can help? chapter 16 forward. pdf/digital version will be much help.
ck new color scheme... ok now lets go on to the bizzzzz

PROGRAMMING TEMPLATE

Classifier system is a kind of genetic algorithms which is mechanically simple. It only involves in generating random number, copying string and exchanging part of string.
Any system whose survival means to respond correctly on inputs from its surroundings can use CS, because what CS does is to classify these inputs and then choose appropriate action based on collection of identified behavior patterns.
These patterns should be frame (knowledge representationally) so as to express them as rules, ie. IF THEN . There could be more than one condition and more than one action, in so far those have logical symbol (AND, NAND, OR, etc). Cross check the rules is important so logic of patterns match, sometimes we need to use # (don't care symbol) to achieve this.

OUR CONTEXT

so which system survival to which? there are three scenarios:
1. space as the system surviving from people/vice versa. this means one CS for space.
2. space and people survive each other. this could mean two CSs, one each for people and space.
3. an outside system which wanting to survive from space and people. this means one outside (people and space) CS. what? yes, a CS which classify both subjects actions and then react. outside system could be driven by observer, or by computer/the CS itself.

from design point of view, i am inclined to identify the context as number 1.
the reason is obvious, any design which can't socially function's not worth the process.
we can implement first level, and perhaps also the second level, of Hillier's generic functions.
context 2 is an ideal model. its like us and our earth type-of-thing, a Godless frame.
context 3 i just thought about it a wee bit, i have to come back to explain this.

KNOWN BEHAVIOR PATTERNS FOR CONTEXT 1

space has only one goal, to enable first level of generic function within. ie. to let people occupy it, move about it and suit intelligibly for them.
1. swarm algorithm means agent is seeing (agents and space), following (to walk to specific target), engaging (with space and also with other agent) which means not moving for sometime.
2. alpha syntax algorithm will need redefinition, because it will act based on what CS told it to do. so, it will generate spaces one at a time, according to agent's action.

CHARACTERS FOR CONTEXT 1

1. an agent; besides following each other (which don't mean they die if they don't follow any agent at any moment), they have to be attaracted to space. attraction to space can be use for enforcing character: die if they are not inside any space.
2. space needs to keep all agents alive all the time; production of space-cell always according to what agents do at any moment.

STORY BOARD FOR CONTEXT 1

1. start with one space-cell.
2. run agents.
3. CS identify agent's action, tell space to grow accordingly.
4. grow more cell/s.
5. back to 3.

Sunday, February 16, 2003

the color plate just not working from laptop view. dont have time to change it now.

ok, starlogo released starlogo 2.0 in november 02, so i try it on. it is much better interface from the older version. and i like the command window where basic codings are presented as titles to choose from. this is advanced :D learning mechanism. it really enable people to work on it from zero. i consider it one stone less educately from w3schools interface, perhaps starlogo team can add "try and see" window in the future?

manage to get tues from p schedule for real avba coding crash lesson. we start next tue.

Tuesday, February 11, 2003

arggggggggggggggghhhhhhhhhhhhhhHHHHHHHHHHHHHHHHHHHHHHHH
feel abit depress today finding swiki site filled with new files........
Chris's classes go on with AVBA, and i missed them all.
well...

howEVER its okay,
anyway, i doubted the whole idea programming with autoCADVBA for my project for the last two months.
what is it with researching??? i DOUBTED everything ........

but now, after looking around for alternatives, first instinct is to go back on that track again.
the thing is, when i was looking around, i feel that what p said on the train that day; about hating - being only offered one mode, about no other way to do things; its quite true.
but hey, what if .. it is the only way?? at least, atm... learn to love what you hate?

ok Chris, i'll be working on my project learning from your class examples...
wish me luck, i need to learn this fast with highest spirit ever

Sunday, February 09, 2003

test ftp'd

Tuesday, February 04, 2003

i called paul last friday, and he advised to redefining the proposal, and the structure of my research. about the proposal, the aim certainly atm had to changed since no philosophical work has been done to identify or even consider an automatable architecture design. i consider this as an issue for me to continue my research in this field, its like the big question in a very big gap practical environment. design architecture automation is like on the opposite far side of what recently considered automatable when just refering to the nature of of it. this is what i got from oct 17, 02 meeting.

about the structure of research, i need to devise what to do now. what i want to do now is to explain what is classifier system to p, because i believe i can use this for my research experiments.

classifier system, as it named is something that classifies. what's it classifies? it classifies inputs from environment to enable itself as a system to learn about the system and how this learning could be use to optimal its existence. ok, for example a frog has to survive in an environment, the original reading is here. what it has to do is to eat the right food by identifying which food available and choose the right one to eat. the right food (ie. something the programmer will have to know preconceptually) is small flying insects

http://www.faqs.org/faqs/ai-faq/genetic/part2/section-5.html