GysdeJongh <jongh711@planet.nl> wrote:
> "Chris Malcolm" <cam@holyrood.ed.ac.uk> wrote in message
> news:4r8lhhFq8dtjU1@individual.net...
>> Will, T2 <wmmckee@cox.net> wrote:
>>> On Sat, 04 Nov 2006 20:29:01 +0000, Trinkwasser
>>> <spam@devnull.com.invalid> wrote:
>> IMHO the really important thing to grasp about diabetes is that you're
>> dealing with a damaged control system.
> Hi Chris Malcolm,
> why >damaged > ??????
It's stopped working properly.
> It is just another control system ???
> With less feedback loops maybe ????
There's an important difference between a properly working control
system and one that isn't.
>>If you're a T2 it still works
>> to some extent, just not well enough. If you're a T1 it's
>> broken. Treating diabetes, whether T1 or T2, isn't just a case of
>> supplying missing stuff or avoiding problematic challenges to the
>> system, you end up interacting with the parts of the system that still
>> work, even if they're not working properly any longer.
> What is >properly > ?????????
> I do not have any book on system theory with the term >properly >
> I do not believe that the term >properly > even has a meaning in science
Aha! I think I see your problem. In descriptive science how on earth
can we possibly justify having purposes as part of the natural world?
Someone has scared you with the bogeyman of improperly teleological
descriptions in science. Quite properly too. Improperly teleological
descriptions have to got rid of. Proper one, however, are not only
fine but essential.
Welcome to engineering, a world in which things working properly is an
important fundamental concept. Welcome to biology. Some physics
enviers did once try to "purify" biology of all teleological
description, not just improper teleology. Fortunately the attempt
failed. Otherwise how could your doctor possibly find out what was
wrong with you without a concept of how your body ought to be behaving
if it all was working properly? It can be done, but it's much more
difficult.
>> It's a dynamic problem in which time and history play important
>> parts. A typical feature of damaged control systems is that when you
>> tinker with them you often get inconsistent paradoxical
>> counter-intuitive responses from the system. In fact getting
> Intuition is nothing more then unconscious knowledge.
Not necessarily unconscious. The ineffable is not necessarily
unconscious. I'd prefer a term like tacit (e.g. as in Polanyi's
"tacit knowledge").
> If it is > counter-intuitive > then you learn from it and the next time the
> same event is not > counter-intuitive > any more.
You're quite right.
>> inconsistent paradoxical responses to your tinkering is a useful clue
>> that what you're dealing with is a damaged control system.
> what is >inconsistent paradoxical responses < ????????
I don't think you can have spent much time in workshops or
laboratories. It's when you keep thinking you have a good idea about
how to make things better and you often make them worse instead, and
you can't understand what the hell is going on. You end up baffled,
almost convinced there must be a malignant demon corrupting your
observations or something. That's inconsistent paradoxical responses.
In medicine, in engineering, and in ecology, that's often a useful
clue that there's a control system involved, or if you already knew
that, that it's doing something significantly different than what you
supposed.
Inconsistent paradoxical responses, for example, often baffle people
in asd who're trying to lower their morning fasting BGs.
If you were trying to do something about indigestion and stomach
acidity and weren't having much success with antacids, and discovered
that the problems disappeared if you drank some lemon juice or
vinegar, that would be a paradoxical response.
Finding that eating a small amount of carb lowers your BG is a
paradoxical response.
An inconsistent paradoxixal response is one that somerimes happens and
sometimes doesn't.
>> Most doctors and medical researchers don't understand control systems
>> because they're not taught about them, and few of them have enough
>> maths and physics to pick it up even if they wanted to.
> I don't think > maths and physics > has anything to do with it
After doing my best to give students a non-mathematical introduction
to the typical behaviours of simple PID control systems, and then
asking them what kind of tinkering they would try in order to fix
various undesirable behaviour by the system, it was those who had a
maths and physics educational background who fared much the best. That
was my observation. It was backed up by a statistical analysis of many
years of exam results, compared to the educational backgrounds of the
students.
> Once I looked up from my keyboard and saw him jump from the roof of my shed
> to the top of the fence to catch a bird.
> The fench is wood , the top is about 1 inch wide.
> I wanted to do the > maths and physics > of this event
> I soon realised that it would take me the rest of the day or , if I wanted
> to include his equilibrium and muscle movements , the rest of the week...
I think you severely underestimate the size of the problem.
> Now was my tomcat a mathematician ???
> Not at all
> Evolution found something far more effective : neural networks
Of course you don't need to use the mathematical model of something as
an ingredient in implementing the behaviour. That's an elementary
tempting error that lots of young engineers and physical scientists
fall into.
> He had a fight
> Long story short : he got an infection in his left eye , it had to be
> removed....
> He got > inconsistent paradoxical responses > when he tried to , slowly ,
> reach his tin of salmon on the kitchen floor.But he was a living creature so
> he learned. He > tinkered > his > damaged control system > he did not
> get > inconsistent paradoxical responses > : the next day he still had
> only one eye. He just re-programmed his neural network.
> I looked up from my keyboard and saw him jump from the roof of my shed to
> the top of the fench.
> No bird this time
> Just for fun I guess
> My message:
> if you keep measuring your bg 1h and 2h postprandial you program your neural
> network.
You're absolutely correct. But I suspect you haven't actually tried to
get neural networks to tinker with control systems in order to adapt
them to new conditions. If you had you would know that while it can be
made to work very well indeed, a most unfortunate charactristic is how
long it takes when more than very simple parametric hill-climbing is
involved. Evolution is also pretty good at ending up with very good
complex designs, like the eye or the wing, even though all it does is
blindly tinker and kill off the mistakes. It too suffers from the
disadvantage of taking a very long time to solve all but the simplest
problems.
That is why evolution, after billions of years, eventually ended up
with intelligence, because intelligent brains are capable of
developing understanding and insight. That works hugely much faster
than the blind tinkering of neural networks.
> Totally unnecessary to make graphs or do math or physics...
> After you have trained your personal neural network you "know" by
> "intuition" what would spike you.....
If certain things often spike you, you'll be able intuitively to learn
that pretty fast, because it's a simple case of identifying simple
cause and effect. In more complicated multivariable problems where
history is also important, such as what you ate yesterday, then if you
have no insight or understanding neural nets that will certainly get there
eventually, *if* you have the patience.
When a complicated and damaged control system is involved, welcome to
the world of paradoxical inconsistent responses which can baffle
understanding and drag out the adaptation of neural nets way beyond
your patience.
--
Chris Malcolm
cam@infirmatics.ed.ac.uk DoD #205
IPAB, Informatics, JCMB, King's Buildings, Edinburgh, EH9 3JZ, UK
[
http://www.dai.ed.ac.uk/homes/cam/] 
Re: Whoe grain competition 
m1nk2tfkli7if45su855h8i05asjfb3aq@4ax.com. ..
Bit expensive though.
Linear Mode
