You're serious, aren't you?

CSGnet Archive CSG1993
1

[From Rick Marken (930617.1200)]

Martin Taylor (930617 13:45) --

We have tried to point out in various ways, starting from differently
phrased but equivalent bases, that failure to perform a reconstruction
was inconclusive about whether the perceptual signal contains information
about the disturbance.

Sounds like a win-win proposition to me.

But tell me, is there some way that I can know in advance whether or
not I am going to succeed at my reconstruction? Not that that's
important or anything. Oh, by the way, did you know that there is
information in a flipped coin about how it's going to land? Yeah. Really.
I know this because I can look at films of coin flips and reconstruct what
the result of each flip was based on the information in the flip. Sometimes
I reconstruct the result correctly -- I say tails when it was tails; and I say
heads when it was heads. Of course, I can't tell what happened after some
of the flips; for example, I predict tails it turns out to be heads. But this
failure to reconstruct the result of the flip based on the information in
the flip is certainly inconclusive about whether there is information in the
flip about how the coin will land; the correct reconstructions prove that
there is information in the flip about how the coin will land.

Oh, and I'm selling my coin flip information detection system for just
$1,000,000 each (nothing compared to the amount you can expect to
make with this powerful tool). If you order now, you can get a
control system perceptual information detector thrown in for FREE.

Despite this, we keep reading that if the
reconstruction is imperfect, then THERE IS NO INFORMATION ABOUT THE
DISTURBANCE IN THE PERCEPTUAL SIGNAL (capitals courtesy of Rick).

Only because I can't italicize and underline too.

Keep enjoying that old time information.

Best

Rick

2

[Martin Taylor 930617 20:30]
(Rick Marken 930617.1200)

After this, I'm going to stop direct responses that seem to me to have
more the character of a private bicker than of a useful public
discussion. I hate 'tis--'tisn't "arguments."

But tell me, is there some way that I can know in advance whether or
not I am going to succeed at my reconstruction? Not that that's
important or anything.

Didn't the answers to your three questions answer that?

Why do you keep focussing on reconstruction? It's the unreconstructible
part of the disturbance that leads to the need for more layers of ECSs
in the hierarchy. If you can control in one layer, why go for more?
The question at issue (for me) is how much is there that needs to be
dealt with by higher layers. The control system can control using whatever
information it gets through the perceptual system. That much can be
reconstructed. What it can't control can't be reconstructed using the
perceptual signal, the reference signal and the output function. The
questions have the same answer.

The important issues are things like perceptual bandwidth and resolution,
and differential system gain at different output levels. Those determine
limiting information rates of control, and determine in part what needs
to be done beyond the simple ECS, whether it be in parallel ECSs or in
higher levels.

Anyway, the general answer is: what is controlled (compensated for) can
be reconstructed using the perceptual signal, the reference, and the
output function.

Since you know "exactly" how a control system works, I presume that tells
you "exactly" what can be reconstructed and what cannot.

Martin