"jittering"

[Bjorn Simonsen (2006.03.13,09:55 EUST)]

Martin Taylor 940207 11:45

A worthy project, but not one I can address this
week. I’m behind in

preparing a presentation for a workshop this
Thursday and a meeting

on a different topic on Friday. An a major report
is due March 31,

which is only partly drafted. So remind me in a week,
or perhaps

better in April, and maybe we can put our
electronic heads together

to see if either way of looking at the process is
sufficiently well

defined to make differential predictions.

I should not disturb you, but I spend time on the
World Model.

I went back to CGS 1994 (February) and missing
knowledge makes problems for me.

I understand the concept “jittering” underlies the
World Model.

Martin stated “Martin
Taylor 940207 11:45” “that there were two causes for a discrepancy between
imagined and real perception: (1) the additive disturbance, and (2) the world transfer function is not at
this moment matched by the world model function. These causes are not distinguishable at any given instant, but as
several people have pointed out, they can be made distinguishable over time, if
the output is jittered (I think of at least Hans Blom, Tom Bourbon, Jeff
Hunter). The nature of the required jitter is discussed at some length in the
book “Self‑tuning systems” that Hans recommended. The reason it
can work is that the output jitter is uncorrelated with the world’s
disturbance, so that on average the discrepancies due to the additive
disturbance will sum to zero when the world model is accurate.”

I have not studied the book “Self-tuning
systems”, and I don’t think I am going to read it. I should have done it
because I have problems with the concept “jittering”. I understand this is a
concept used in IT. It states the difference between the effect of not using
the information about a certain value but using the value a certain GIS uses
(the technology).

You don’t need to comment your statement from
1994, but if you find time, answer me if my understanding of your use of “jittering”
is useless.

bjorn

···

Re: “jittering”
[Martin Taylor 2006.03.13.16.44]

[Bjorn Simonsen (2006.03.13,09:55
EUST)]
Martin Taylor 940207 11:45

I spend time on the World
Model.
I went back to CGS 1994 (February) and missing
knowledge makes problems for me.

Wow! How do you find the relevant stuff in among those thousands
of messages?

I understand the concept “jittering” underlies the
World Model.

Not really. The bit you quote only argues for it as one way the
internal part of a control system can assess the properties of the
environmental feeback function independently of any regularities in
the disturbance. If the “World Model” is to substitute for
the external environment when we imagine the effects of our actions,
jittering the output when we are controlling through the real world
can help. It doesn’t help when we are imagining (taking the
“through the World Model” view of control in
imagination).

Martin stated “Martin Taylor 940207 11:45” “that
there were two causes for a discrepancy between imagined and real
perception: (1) the additive disturbance, and (2) the world transfer
function is not at this moment matched by the world model
function. These causes are not distinguishable at any given
instant, but as several people have pointed out, they can be made
distinguishable over time, if the output is jittered (I think of at
least Hans Blom, Tom Bourbon, Jeff Hunter). The nature of the required
jitter is discussed at some length in the book “Self-tuning
systems” that Hans recommended. The reason it can work is that
the output jitter is uncorrelated with the world’s disturbance, so
that on average the discrepancies due to the additive disturbance will
sum to zero when the world model is accurate.”

I have not
studied the book “Self-tuning systems”, and I don’t think I am
going to read it. I should have done it because I have problems with
the concept “jittering”. I understand this is a concept used in
IT.

IT = Information Technology? If so, how is it used there?

It states the difference between the effect of
not using the information about a certain value but using the value a
certain GIS uses (the technology).

I don’t understand this.

In the quoted paragraph from 1994, the idea of “jittering”
is to produce output that is NOT related to the disturbance, to see
what effect it has on the perception. If you do this, you can use any
kind of output waveform, whereas if you simply do your best to
control, you are limited to waveforms that, after passing through the
environmental feedback path, oppose the disturbance waveform. That
gives you less opportunity to assess how the world actually behaves,
and means that it might be harder to control against different
disturbance patterns in the future. It’s an aspect of reorganization
that isn’t often talked about (maybe not since 1994:-).

You don’t need to comment your statement from 1994,
but if you find time, answer me if my understanding of your use of
“jittering” is useless.

I didn’t
really get what your understanding is, but I hope what I wrote will be
enough to let you know whether your understanding was
correct.

I don’t
think I will probably respond much more this week.

Martin