Control by Nervous System (was Re: Are academics stuck...)

rsmarken · August 14, 2017, 4:19pm

[From Rick Marken (2017.08.14.0920)]

···

Rupert Young (2017.08.13 17.00)–

  RY: I refer you to the paper on p170 of this large document,
“Simulation of Human Balance Control Using an Inverted,Pendulum
Model,” and offer for your delectation (and despair) the comments in the
abstract,Â

    "It is probable that there are many concurrent control loops
occurring in the central nervous system"Â and,
“no models of how these controllers might operate within the
nervous system have yet been developed.”

Â

RY: Below is my email to the authors…

Â

    RY: Actually there is a very detailed model of the functional
control architecture of the nervous system, which you can see in
the book “Behavior: The Control Of Perception”

Â

RM: There is one fundamental question about how control systems are implemented in the nervous system that, to my knowledge, has not been confirmed by neurophysiological observation. That is the question of how the nervous system handles the fact that control requires that error signals be both positive and negative.Â

RM: The PCT model of control by the nervous system assumes that control signals – perceptual, reference, error and output – are carried as neural firing rates, which can only be positive. So how does the nervous system implement the “comparator” function, r - p?Â If p < r then there is no problem; the output of this function, the error signal, is a positive neural firing rate. However, if p > r then the output of this function is negative and there is no such thing as a negative rate of neural firing.Â

RM: The obvious solution to this problem is to have control systems be made up of two parallel comparators that receive the same perceptual and reference signal inputs but do the comparison function in the opposite ways and produces separate output (error) signals; one comparator function is r-p and the other is p - r. When p < r the r-p comparator produces an error signal while the p-r comparator output remains at 0; when p > r, the p-r comparator produces an error signal while the r - p comparator remains at 0. The outputs of each of these comparators are always positive so they must be connected to the output function so that the output of the r-p comparator excites and the output of the p-r comparator inhibits output.Â

RM: I am not aware of any neurophysiological evidence for the existence of such a “parallel comparator” architecture that would support the existence of control systems in the nervous system. It seems like this is one of the most basic implications of the control model for the neurophysiological basis of perceptual control. So while Â the “parallel comparator” system does provide a model of how control systems “might operate within the nervous system”, it doesn’t prove that they do operate this way. Since it seems that there are some neurophysiologists on CSGNet I wonder if they know of any evidence for the existence of the kind of “parallel comparator” neural architecture required to implement control systems in the nervous system.Â

BestÂ

Rick

–

Richard S. MarkenÂ

"Perfection is achieved not when you have nothing more to add, but when you
have nothing left to take away.â€?
Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â --Antoine de Saint-Exupery

MartinT · August 14, 2017, 5:51pm

[Martin Taylor 2017.08.14.13.41]

[From Rick Marken (2017.08.14.0920)]

I second this motion! That issue has always been lurking in the

background. The argument has always been from necessity, that since
we can observe the behavioural manifestations of control from either
side of an apparent reference value of a CEV in the environment,
something in the brain must treat both directions of error more or
less the same. Energetically, it would be very wasteful for zero
error in every control unit to be signalled by a firing rate about
midway between zero and the maximum possible, so the balance
presumably isn’t done that way. It would be really nice to know the
mechanism or mechanisms that allow this to happen. We have opposed
musculature that does the physical job in some cases. We can see
that. But is there anything equivalent in the nervous system?

Martin

···

…

      RM: There is one fundamental question about how control

systems are implemented in the nervous system that, to my
knowledge, has not been confirmed by neurophysiological
observation. That is the question of how the nervous system
handles the fact that control requires that error signals be
both positive and negative.

      RM: The PCT model of control by the

nervous system assumes that control signals – perceptual,
reference, error and output – are carried as neural firing
rates, which can only be positive. So how does the nervous
system implement the “comparator” function, r - p? If p <
r then there is no problem; the output of this function, the
error signal, is a positive neural firing rate. However, if p

r then the output of this function is negative and there
is no such thing as a negative rate of neural firing.

      RM: The obvious solution to this

problem is to have control systems be made up of two parallel
comparators that receive the same perceptual and reference
signal inputs but do the comparison function in the opposite
ways and produces separate output (error) signals; one
comparator function is r-p and the other is p - r. When p <
r the r-p comparator produces an error signal while the p-r
comparator output remains at 0; when p > r, the p-r
comparator produces an error signal while the r - p comparator
remains at 0. The outputs of each of these comparators are
always positive so they must be connected to the output
function so that the output of the r-p comparator excites and
the output of the p-r comparator inhibits output.

      RM: I am not aware of any

neurophysiological evidence for the existence of such a
“parallel comparator” architecture that would support the
existence of control systems in the nervous system. It seems
like this is one of the most basic implications of the control
model for the neurophysiological basis of perceptual control.
So while the “parallel comparator” system does provide a
model of how control systems “might operate within the nervous
system”, it doesn’t prove that they do operate this way. Since
it seems that there are some neurophysiologists on CSGNet I
wonder if they know of any evidence for the existence of the
kind of “parallel comparator” neural architecture required to
implement control systems in the nervous system.

FredNickols · August 14, 2017, 6:28pm

[From Fred Nickols (2017.08.14.1426 ET)]

I’m probably way out of my depth here but I don’t know that the comparator function needs to assign polarity (i.e., negative or positive). What if all the comparator detects is the magnitude of the error? Then it wouldn’t matter which was greater than the other.

Fred Nickols

···

From: Martin Taylor [mailto:mmt-csg@mmtaylor.net]
Sent: Monday, August 14, 2017 1:52 PM
To: csgnet@lists.illinois.edu
Subject: Re: Control by Nervous System (was Re: Are academics stuck…)

[Martin Taylor 2017.08.14.13.41]

[From Rick Marken (2017.08.14.0920)]

…

RM: There is one fundamental question about how control systems are implemented in the nervous system that, to my knowledge, has not been confirmed by neurophysiological observation. That is the question of how the nervous system handles the fact that control requires that error signals be both positive and negative.

RM: The PCT model of control by the nervous system assumes that control signals – perceptual, reference, error and output – are carried as neural firing rates, which can only be positive. So how does the nervous system implement the “comparator” function, r - p? If p < r then there is no problem; the output of this function, the error signal, is a positive neural firing rate. However, if p > r then the output of this function is negative and there is no such thing as a negative rate of neural firing.

RM: The obvious solution to this problem is to have control systems be made up of two parallel comparators that receive the same perceptual and reference signal inputs but do the comparison function in the opposite ways and produces separate output (error) signals; one comparator function is r-p and the other is p - r. When p < r the r-p comparator produces an error signal while the p-r comparator output remains at 0; when p > r, the p-r comparator produces an error signal while the r - p comparator remains at 0. The outputs of each of these comparators are always positive so they must be connected to the output function so that the output of the r-p comparator excites and the output of the p-r comparator inhibits output.

RM: I am not aware of any neurophysiological evidence for the existence of such a “parallel comparator” architecture that would support the existence of control systems in the nervous system. It seems like this is one of the most basic implications of the control model for the neurophysiological basis of perceptual control. So while the “parallel comparator” system does provide a model of how control systems “might operate within the nervous system”, it doesn’t prove that they do operate this way. Since it seems that there are some neurophysiologists on CSGNet I wonder if they know of any evidence for the existence of the kind of “parallel comparator” neural architecture required to implement control systems in the nervous system.

I second this motion! That issue has always been lurking in the background. The argument has always been from necessity, that since we can observe the behavioural manifestations of control from either side of an apparent reference value of a CEV in the environment, something in the brain must treat both directions of error more or less the same. Energetically, it would be very wasteful for zero error in every control unit to be signalled by a firing rate about midway between zero and the maximum possible, so the balance presumably isn’t done that way. It would be really nice to know the mechanism or mechanisms that allow this to happen. We have opposed musculature that does the physical job in some cases. We can see that. But is there anything equivalent in the nervous system?

Martin

rsmarken · August 14, 2017, 7:45pm

[From Rick Marken (2017.08.14. 1245)]

Fred Nickols (2017.08.14.1426 ET)--

Â

FN: Iâ€™m probably way out of my depth here but I donâ€™t know that the comparator function needs to assign polarity (i.e., negative or positive).Â What if all the comparator detects is the magnitude of the error? Then it wouldnâ€™t matter which was greater than the other.

RM: This is equivalent to saying that the comparator function is |r-p|, the absolute value of the difference between r and p. The problem with this is that, when p is greater than the reference, the output of the comparator must drive system output so as to reduce p in proportion to the degree to which it exceeds r; that is, the feedback effects of output on input must be negative. If the error signal produced by the comparator were proportional only to the magnitude of the difference between p and r (rather than to its direction -- greater than or less than r -- as well) , then the feedback would not be negative when p exceeded r and there would be no control in that range of values of p.Â
BestÂ
Rick
Â >

···

Â

Fred Nickols

Â

From: Martin Taylor [mailto:mmt-csg@mmtaylor.net>mmt-csg@mmtaylor.net]
Sent: Monday, August 14, 2017 1:52 PM
To: <mailto:csgnet@lists.illinois.edu>csgnet@lists.illinois.edu
Subject: Re: Control by Nervous System (was Re: Are academics stuck...)

Â

[Martin Taylor 2017.08.14.13.41]

[From Rick Marken (2017.08.14.0920)]

...

RM: There is one fundamental question about how control systems are implemented in the nervous system that, to my knowledge, has not been confirmed by neurophysiological observation. That is the question of how the nervous system handles the fact that control requires that error signals be both positive and negative.Â

Â

RM: The PCT model of control by the nervous system assumes that control signals -- perceptual, reference, error and output -- are carried as neural firing rates, which can only be positive. So how does the nervous system implement the "comparator" function, r - p?Â If p < r then there is no problem; the output of this function, the error signal, is a positive neural firing rate. However, if p > r then the output of this function is negative and there is no such thing as a negative rate of neural firing.Â

Â

RM: The obvious solution to this problem is to have control systems be made up of two parallel comparators that receive the same perceptual and reference signal inputs but do the comparison function in the opposite ways and produces separate output (error) signals; one comparator function is r-p and the other is p - r. When p < r the r-p comparator produces an error signal while the p-r comparator output remains at 0; when p > r, the p-r comparator produces an error signal while the r - p comparator remains at 0. The outputs of each of these comparators are always positive so they must be connected to the output function so that the output of the r-p comparator excites and the output of the p-r comparator inhibits output.Â

Â

RM: I am not aware of any neurophysiological evidence for the existence of such a "parallel comparator" architecture that would support the existence of control systems in the nervous system. It seems like this is one of the most basic implications of the control model for the neurophysiological basis of perceptual control. So while Â the "parallel comparator" system does provide a model of how control systems "might operate within the nervous system", it doesn't prove that they do operate this way. Since it seems that there are some neurophysiologists on CSGNet I wonder if they know of any evidence for the existence of the kind of "parallel comparator" neural architecture required to implement control systems in the nervous system.Â

Â

I second this motion! That issue has always been lurking in the background. The argument has always been from necessity, that since we can observe the behavioural manifestations of control from either side of an apparent reference value of a CEV in the environment, something in the brain must treat both directions of error more or less the same. Energetically, it would be very wasteful for zero error in every control unit to be signalled by a firing rate about midway between zero and the maximum possible, so the balance presumably isn't done that way. It would be really nice to know the mechanism or mechanisms that allow this to happen. We have opposed musculature that does the physical job in some cases. We can see that. But is there anything equivalent in the nervous system?

Martin

--
Richard S. MarkenÂ
"Perfection is achieved not when you have nothing more to add, but when you
have nothing left to take away.â€?
Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â --Antoine de Saint-Exupery

FredNickols · August 14, 2017, 8:31pm

[From Fred Nickols (2017.08.14.1630 ET)]

If Iâ€™m trying to maintain my car in the center of the lane, what if Iâ€™ve got one system that controls leftward deviation and one that controls rightward deviation?Â Then magnitude of the error signal would be all that matters wouldnâ€™t it?

Fred

···

From: Richard Marken [mailto:rsmarken@gmail.com]
Sent: Monday, August 14, 2017 3:46 PM
To: csgnet@lists.illinois.edu
Subject: Re: Control by Nervous System (was Re: Are academics stuck…)

[From Rick Marken (2017.08.14. 1245)]

Fred Nickols (2017.08.14.1426 ET)–

FN: Iâ€™m probably way out of my depth here but I donâ€™t know that the comparator function needs to assign polarity (i.e., negative or positive). What if all the comparator detects is the magnitude of the error? Then it wouldnâ€™t matter which was greater than the other.

RM: This is equivalent to saying that the comparator function is |r-p|, the absolute value of the difference between r and p. The problem with this is that, when p is greater than the reference, the output of the comparator must drive system output so as to reduce p in proportion to the degree to which it exceeds r; that is, the feedback effects of output on input must be negative. If the error signal produced by the comparator were proportional only to the magnitude of the difference between p and r (rather than to its direction – greater than or less than r – as well) , then the feedback would not be negative when p exceeded r and there would be no control in that range of values of p.

Best

Rick

Fred Nickols

From: Martin Taylor [mailto:mmt-csg@mmtaylor.net]
Sent: Monday, August 14, 2017 1:52 PM
To: csgnet@lists.illinois.edu
Subject: Re: Control by Nervous System (was Re: Are academics stuck…)

[Martin Taylor 2017.08.14.13.41]

[From Rick Marken (2017.08.14.0920)]

…

RM: There is one fundamental question about how control systems are implemented in the nervous system that, to my knowledge, has not been confirmed by neurophysiological observation. That is the question of how the nervous system handles the fact that control requires that error signals be both positive and negative.

RM: The PCT model of control by the nervous system assumes that control signals – perceptual, reference, error and output – are carried as neural firing rates, which can only be positive. So how does the nervous system implement the “comparator” function, r - p? If p < r then there is no problem; the output of this function, the error signal, is a positive neural firing rate. However, if p > r then the output of this function is negative and there is no such thing as a negative rate of neural firing.

RM: The obvious solution to this problem is to have control systems be made up of two parallel comparators that receive the same perceptual and reference signal inputs but do the comparison function in the opposite ways and produces separate output (error) signals; one comparator function is r-p and the other is p - r. When p < r the r-p comparator produces an error signal while the p-r comparator output remains at 0; when p > r, the p-r comparator produces an error signal while the r - p comparator remains at 0. The outputs of each of these comparators are always positive so they must be connected to the output function so that the output of the r-p comparator excites and the output of the p-r comparator inhibits output.

RM: I am not aware of any neurophysiological evidence for the existence of such a “parallel comparator” architecture that would support the existence of control systems in the nervous system. It seems like this is one of the most basic implications of the control model for the neurophysiological basis of perceptual control. So while the “parallel comparator” system does provide a model of how control systems “might operate within the nervous system”, it doesn’t prove that they do operate this way. Since it seems that there are some neurophysiologists on CSGNet I wonder if they know of any evidence for the existence of the kind of “parallel comparator” neural architecture required to implement control systems in the nervous system.

I second this motion! That issue has always been lurking in the background. The argument has always been from necessity, that since we can observe the behavioural manifestations of control from either side of an apparent reference value of a CEV in the environment, something in the brain must treat both directions of error more or less the same. Energetically, it would be very wasteful for zero error in every control unit to be signalled by a firing rate about midway between zero and the maximum possible, so the balance presumably isn’t done that way. It would be really nice to know the mechanism or mechanisms that allow this to happen. We have opposed musculature that does the physical job in some cases. We can see that. But is there anything equivalent in the nervous system?

Martin

–

Richard S. Marken

"Perfection is achieved not when you have nothing more to add, but when you
have nothing left to take away.â€?
–Antoine de Saint-Exupery

MartinT · August 14, 2017, 9:03pm

[From Fred Nickols (2017.08.14.1630 ET)]

Â

      If Iâ€™m trying to maintain my car in the

center of the lane, what if Iâ€™ve got one system that controls
leftward deviation and one that controls rightward deviation?Â
Then magnitude of the error signal would be all that matters
wouldnâ€™t it?

Â

Fred

···

[From Rick Marken (2017.08.14. 1245)]

Â

                  Fred
Nickols (2017.08.14.1426 ET)–

Â

                  FN:
Iâ€™m probably way out of my depth here but I donâ€™t
know that the comparator function needs to assign
polarity (i.e., negative or positive).Â What if
all the comparator detects is the magnitude of the
error? Then it wouldnâ€™t matter which was greater
than the other.

Â

              RM: This is equivalent to saying

that the comparator function is |r-p|, the absolute
value of the difference between r and p. The problem
with this is that, when p is greater than the
reference, the output of the comparator must drive
system output so as to reduce p in proportion to the
degree to which it exceeds r; that is, the feedback
effects of output on input must be negative. If the
error signal produced by the comparator were
proportional only to the magnitude of the difference
between p and r (rather than to its direction –
greater than or less than r – as well) , then the
feedback would not be negative when p exceeded r and
there would be no control in that range of values of
p.Â

Â

BestÂ

Â

Rick

Â

Â

                  Fred

Nickols

Â

From:
Martin Taylor [mailto:mmt-csg@mmtaylor.net ]
Sent: Monday, August 14, 2017 1:52 PM
To: csgnet@lists.illinois.edu
Subject: Re: Control by Nervous System
(was Re: Are academics stuck…)

Â

[Martin Taylor 2017.08.14.13.41]

                          [From

Rick Marken (2017.08.14.0920)]

...

                            RM:
There is one fundamental question about
how control systems are implemented in
the nervous system that, to my
knowledge, has not been confirmed by
neurophysiological observation. That is
the question of how the nervous system
handles the fact that control requires
that error signals be both positive and
negative.Â

Â

                            RM:
The PCT model of control by the nervous
system assumes that control signals –
perceptual, reference, error and output
– are carried as neural firing rates,
which can only be positive. So how does
the nervous system implement the
“comparator” function, r - p?Â If p <
r then there is no problem; the output
of this function, the error signal, is a
positive neural firing rate. However, if
p > r then the output of this
function is negative and there is no
such thing as a negative rate of neural
firing.Â

Â

                            RM:
The obvious solution to this problem is
to have control systems be made up of
two parallel comparators that receive
the same perceptual and reference signal
inputs but do the comparison function in
the opposite ways and produces separate
output (error) signals; one comparator
function is r-p and the other is p - r.
When p < r the r-p comparator
produces an error signal while the p-r
comparator output remains at 0; when p

r, the p-r comparator produces an
error signal while the r - p comparator
remains at 0. The outputs of each of
these comparators are always positive so
they must be connected to the output
function so that the output of the r-p
comparator excites and the output of the
p-r comparator inhibits output.Â

Â

                            RM:
I am not aware of any neurophysiological
evidence for the existence of such a
“parallel comparator” architecture that
would support the existence of control
systems in the nervous system. It seems
like this is one of the most basic
implications of the control model for
the neurophysiological basis of
perceptual control. So while Â the
“parallel comparator” system does
provide a model of how control systems
“might operate within the nervous
system”, it doesn’t prove that they do
operate this way. Since it seems that
there are some neurophysiologists on
CSGNet I wonder if they know of any
evidence for the existence of the kind
of “parallel comparator” neural
architecture required to implement
control systems in the nervous system.Â

Â

                      I
second this motion! That issue has always been
lurking in the background. The argument has
always been from necessity, that since we can
observe the behavioural manifestations of
control from either side of an apparent
reference value of a CEV in the environment,
something in the brain must treat both
directions of error more or less the same.
Energetically, it would be very wasteful for
zero error in every control unit to be
signalled by a firing rate about midway
between zero and the maximum possible, so the
balance presumably isn’t done that way. It
would be really nice to know the mechanism or
mechanisms that allow this to happen. We have
opposed musculature that does the physical job
in some cases. We can see that. But is there
anything equivalent in the nervous system?
                      Martin

Â

–

                                  Richard S.

MarkenÂ

                                    "Perfection

is achieved not when you have
nothing more to add, but when
you

                                    have nothing left to take away.â€?

                                    Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â

–Antoine de Saint-Exupery

FredNickols · August 14, 2017, 9:27pm

[From Fred Nickols (2017.08.14.1725 ET)]

Gee, thanks, Martin.Â I guess Iâ€™m not so stupid after all.Â Thanks for stating my proposition more clearly than I did.

One question: Why does the magnitude of the error signal not matter.Â If itâ€™s a big leftward deviation and it takes a big change in steering to reduce that error, doesnâ€™t magnitude matter?

Fred Nickols

···

From: Martin Taylor [mailto:mmt-csg@mmtaylor.net]
Sent: Monday, August 14, 2017 5:04 PM
To: csgnet@lists.illinois.edu
Subject: Re: Control by Nervous System (was Re: Are academics stuck…)

[Martin Taylor 2017.08.14.16.45]

[From Fred Nickols (2017.08.14.1630 ET)]

If Iâ€™m trying to maintain my car in the center of the lane, what if Iâ€™ve got one system that controls leftward deviation and one that controls rightward deviation? Then magnitude of the error signal would be all that matters wouldnâ€™t it?

Fred

If I understand your proposition, if you deviated leftward, the left-controlling system would have error but the right-controller would not. So the left system would pull the car back toward the centre, while the right system would do nothing. If you deviated to the right, the opposite would be true. The two one-way systems work together like one two-way system. What matters is not the magnitude of the error signal, but the difference between the error signals in the two systems.

At least that’s the basic function. You can get a lot more out of such a system if you can bias the zero-error point where perception equals reference of each one-way system. If the left system’s zero error doesn’t happen until there is a little rightward deviation, and vice-versa, when the overall error is zero, the two systems oppose one another, like the two muscles that make a joint angle move one way or the other. With overlap, you get a “stiff” control region because the two systems are in Kent McClelland’s conflictive collective control state, where the apparent loop gain is the sum of the individual gains. If the “zero-error” bias is the other way, so there’s a region over which neither system produces an error signal, you get a tolerance zone.

I don’t know whether any of this actually happens in a nervous system, but as soon as you talk about opposed one-way control systems, these possibilities become things to be asked about.

Martin

From: Richard Marken [mailto:rsmarken@gmail.com]
Sent: Monday, August 14, 2017 3:46 PM
To: csgnet@lists.illinois.edu
Subject: Re: Control by Nervous System (was Re: Are academics stuck…)

[From Rick Marken (2017.08.14. 1245)]

Fred Nickols (2017.08.14.1426 ET)–

FN: Iâ€™m probably way out of my depth here but I donâ€™t know that the comparator function needs to assign polarity (i.e., negative or positive). What if all the comparator detects is the magnitude of the error? Then it wouldnâ€™t matter which was greater than the other.

RM: This is equivalent to saying that the comparator function is |r-p|, the absolute value of the difference between r and p. The problem with this is that, when p is greater than the reference, the output of the comparator must drive system output so as to reduce p in proportion to the degree to which it exceeds r; that is, the feedback effects of output on input must be negative. If the error signal produced by the comparator were proportional only to the magnitude of the difference between p and r (rather than to its direction – greater than or less than r – as well) , then the feedback would not be negative when p exceeded r and there would be no control in that range of values of p.

Best

Rick

Fred Nickols

From: Martin Taylor [mailto:mmt-csg@mmtaylor.net]
Sent: Monday, August 14, 2017 1:52 PM
To: csgnet@lists.illinois.edu
Subject: Re: Control by Nervous System (was Re: Are academics stuck…)

[Martin Taylor 2017.08.14.13.41]

[From Rick Marken (2017.08.14.0920)]

…

RM: There is one fundamental question about how control systems are implemented in the nervous system that, to my knowledge, has not been confirmed by neurophysiological observation. That is the question of how the nervous system handles the fact that control requires that error signals be both positive and negative.

RM: The PCT model of control by the nervous system assumes that control signals – perceptual, reference, error and output – are carried as neural firing rates, which can only be positive. So how does the nervous system implement the “comparator” function, r - p? If p < r then there is no problem; the output of this function, the error signal, is a positive neural firing rate. However, if p > r then the output of this function is negative and there is no such thing as a negative rate of neural firing.

RM: The obvious solution to this problem is to have control systems be made up of two parallel comparators that receive the same perceptual and reference signal inputs but do the comparison function in the opposite ways and produces separate output (error) signals; one comparator function is r-p and the other is p - r. When p < r the r-p comparator produces an error signal while the p-r comparator output remains at 0; when p > r, the p-r comparator produces an error signal while the r - p comparator remains at 0. The outputs of each of these comparators are always positive so they must be connected to the output function so that the output of the r-p comparator excites and the output of the p-r comparator inhibits output.

RM: I am not aware of any neurophysiological evidence for the existence of such a “parallel comparator” architecture that would support the existence of control systems in the nervous system. It seems like this is one of the most basic implications of the control model for the neurophysiological basis of perceptual control. So while the “parallel comparator” system does provide a model of how control systems “might operate within the nervous system”, it doesn’t prove that they do operate this way. Since it seems that there are some neurophysiologists on CSGNet I wonder if they know of any evidence for the existence of the kind of “parallel comparator” neural architecture required to implement control systems in the nervous system.

I second this motion! That issue has always been lurking in the background. The argument has always been from necessity, that since we can observe the behavioural manifestations of control from either side of an apparent reference value of a CEV in the environment, something in the brain must treat both directions of error more or less the same. Energetically, it would be very wasteful for zero error in every control unit to be signalled by a firing rate about midway between zero and the maximum possible, so the balance presumably isn’t done that way. It would be really nice to know the mechanism or mechanisms that allow this to happen. We have opposed musculature that does the physical job in some cases. We can see that. But is there anything equivalent in the nervous system?

Martin

–

Richard S. Marken

"Perfection is achieved not when you have nothing more to add, but when you
have nothing left to take away.â€?
–Antoine de Saint-Exupery

MartinT · August 15, 2017, 3:45am

[From Fred
Nickols (2017.08.14.1725 ET)]

Â

        Gee, thanks,

Martin.Â I guess Iâ€™m not so stupid after all.Â Thanks for
stating my proposition more clearly than I did.

Â

One

question: Why does the magnitude of the error signal not
matter.Â If itâ€™s a big leftward deviation and it takes a big
change in steering to reduce that error, doesnâ€™t magnitude
matter?

···

From: Martin Taylor
[mailto:mmt-csg@mmtaylor.net]
Sent: Monday, August 14, 2017 5:04 PM
To: csgnet@lists.illinois.edu
Subject: Re: Control by Nervous System (was Re:
Are academics stuck…)

Â

[Martin Taylor 2017.08.14.16.45]

[From Fred Nickols (2017.08.14.1630 ET)]

Â

        If Iâ€™m trying to maintain my car in the
center of the lane, what if Iâ€™ve got one system that
controls leftward deviation and one that controls rightward
deviation?Â Then magnitude of the error signal would be all
that matters wouldnâ€™t it?

Â

Fred

      If I understand your proposition, if you deviated leftward,

the left-controlling system would have error but the
right-controller would not. So the left system would pull the
car back toward the centre, while the right system would do
nothing. If you deviated to the right, the opposite would be
true. The two one-way systems work together like one two-way
system. What matters is not the magnitude of the error signal,
but the difference between the error signals in the two
systems.

      At least that's the basic function. You can get a lot more out

of such a system if you can bias the zero-error point where
perception equals reference of each one-way system. If the
left system’s zero error doesn’t happen until there is a
little rightward deviation, and vice-versa, when the overall
error is zero, the two systems oppose one another, like the
two muscles that make a joint angle move one way or the other.
With overlap, you get a “stiff” control region because the two
systems are in Kent McClelland’s conflictive collective
control state, where the apparent loop gain is the sum of the
individual gains. If the “zero-error” bias is the other way,
so there’s a region over which neither system produces an
error signal, you get a tolerance zone.

      I don't know whether any of this actually happens in a nervous

system, but as soon as you talk about opposed one-way control
systems, these possibilities become things to be asked about.

      Martin

Â

From: Richard Marken [mailto:rsmarken@gmail.com ]
Sent: Monday, August 14, 2017 3:46 PM
To: csgnet@lists.illinois.edu
Subject: Re: Control by Nervous System (was Re: Are
academics stuck…)

Â

              [From Rick Marken (2017.08.14.

1245)]

Â

                    Fred

Nickols (2017.08.14.1426 ET)–

Â

                    FN:
Iâ€™m probably way out of my depth here but I
donâ€™t know that the comparator function needs to
assign polarity (i.e., negative or positive).Â
What if all the comparator detects is the
magnitude of the error? Then it wouldnâ€™t matter
which was greater than the other.

Â

                RM: This is equivalent to saying
that the comparator function is |r-p|, the absolute
value of the difference between r and p. The problem
with this is that, when p is greater than the
reference, the output of the comparator must drive
system output so as to reduce p in proportion to the
degree to which it exceeds r; that is, the feedback
effects of output on input must be negative. If the
error signal produced by the comparator were
proportional only to the magnitude of the difference
between p and r (rather than to its direction –
greater than or less than r – as well) , then the
feedback would not be negative when p exceeded r and
there would be no control in that range of values of
p.Â

Â

BestÂ

Â

Rick

Â

                    Fred

Nickols

Â

From:
Martin Taylor [mailto:mmt-csg@mmtaylor.net ]
Sent: Monday, August 14, 2017 1:52 PM
To: csgnet@lists.illinois.edu
Subject: Re: Control by Nervous
System (was Re: Are academics stuck…)

Â

[Martin Taylor 2017.08.14.13.41]

                            [From

Rick Marken (2017.08.14.0920)]

...

                              RM:
There is one fundamental question
about how control systems are
implemented in the nervous system
that, to my knowledge, has not been
confirmed by neurophysiological
observation. That is the question of
how the nervous system handles the
fact that control requires that error
signals be both positive and
negative.Â

Â

                              RM:
The PCT model of control by the
nervous system assumes that control
signals – perceptual, reference,
error and output – are carried as
neural firing rates, which can only be
positive. So how does the nervous
system implement the “comparator”
function, r - p?Â If p < r then
there is no problem; the output of
this function, the error signal, is a
positive neural firing rate. However,
if p > r then the output of this
function is negative and there is no
such thing as a negative rate of
neural firing.Â

Â

                              RM:
The obvious solution to this problem
is to have control systems be made up
of two parallel comparators that
receive the same perceptual and
reference signal inputs but do the
comparison function in the opposite
ways and produces separate output
(error) signals; one comparator
function is r-p and the other is p -
r. When p < r the r-p comparator
produces an error signal while the p-r
comparator output remains at 0; when p

r, the p-r comparator produces an
error signal while the r - p
comparator remains at 0. The outputs
of each of these comparators are
always positive so they must be
connected to the output function so
that the output of the r-p comparator
excites and the output of the p-r
comparator inhibits output.Â

Â

                              RM:
I am not aware of any
neurophysiological evidence for the
existence of such a “parallel
comparator” architecture that would
support the existence of control
systems in the nervous system. It
seems like this is one of the most
basic implications of the control
model for the neurophysiological basis
of perceptual control. So while Â the
“parallel comparator” system does
provide a model of how control systems
“might operate within the nervous
system”, it doesn’t prove that they do
operate this way. Since it seems that
there are some neurophysiologists on
CSGNet I wonder if they know of any
evidence for the existence of the kind
of “parallel comparator” neural
architecture required to implement
control systems in the nervous
system.Â

Â

                        I
second this motion! That issue has always
been lurking in the background. The argument
has always been from necessity, that since
we can observe the behavioural
manifestations of control from either side
of an apparent reference value of a CEV in
the environment, something in the brain must
treat both directions of error more or less
the same. Energetically, it would be very
wasteful for zero error in every control
unit to be signalled by a firing rate about
midway between zero and the maximum
possible, so the balance presumably isn’t
done that way. It would be really nice to
know the mechanism or mechanisms that allow
this to happen. We have opposed musculature
that does the physical job in some cases. We
can see that. But is there anything
equivalent in the nervous system?
                        Martin

Â

–

                                    Richard S.

MarkenÂ

                                      "Perfection

is achieved not when you have
nothing more to add, but when
you

                                      have nothing left to take

away.â€?

                                      Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â

Â --Antoine de Saint-Exupery

Â

FredNickols · August 15, 2017, 10:34am

[From Fred Nickols (2017.08.15.0631)]

Iâ€™m puzzled by the thermostat example in your second paragraph, Martin, the one referring to turning on the furnace or air conditioner a little or a lot.

I thought furnaces and air conditioners were simply on or off and the hot air or cold air was essentially a fixed temperature.Â Thus, the way they raise or lower the room temperature has to do mainly with how long theyâ€™re on or off, not how cold or hot their output is.Â Hence, your statement in the first paragraph that all the thermostat can do is turn the furnace or air conditioner on or off.

Am I missing something?

Fred Nickols

···

From: Martin Taylor [mailto:mmt-csg@mmtaylor.net]
Sent: Monday, August 14, 2017 11:46 PM
To: csgnet@lists.illinois.edu
Subject: Re: Control by Nervous System (was Re: Are academics stuck…)

[Martin Taylor 2017.08.14.23.33]

[From Fred Nickols (2017.08.14.1725 ET)]

Gee, thanks, Martin. I guess Iâ€™m not so stupid after all. Thanks for stating my proposition more clearly than I did.

One question: Why does the magnitude of the error signal not matter. If itâ€™s a big leftward deviation and it takes a big change in steering to reduce that error, doesnâ€™t magnitude matter?

I didn’t mean to imply that the error magnitude doesn’t matter. In the kinds of tracking systems we often simulate, it does matter. In an ordinary thermostat, it doesn’t. All the thermostat can do is turn the furnace or air conditioner on or off. If it turns both on together, it wastes energy. If there is a temperature range in which neither is on, the temperature can vary freely over this range.

Whether the magnitude of the error matters depends on what you are trying to control and what you can do about it. You could have a tracking system in which it doesn’t matter (give the subject only a switch to move the cursor up, don’t move, or down, and you could have a thermostat in which it does matter: turn the furnace or air conditioner on a little or a lot depending on how far the temperature is from its current set point.

Martin

Fred Nickols

From: Martin Taylor [mailto:mmt-csg@mmtaylor.net]
Sent: Monday, August 14, 2017 5:04 PM
To: csgnet@lists.illinois.edu
Subject: Re: Control by Nervous System (was Re: Are academics stuck…)

[Martin Taylor 2017.08.14.16.45]

[From Fred Nickols (2017.08.14.1630 ET)]

If Iâ€™m trying to maintain my car in the center of the lane, what if Iâ€™ve got one system that controls leftward deviation and one that controls rightward deviation? Then magnitude of the error signal would be all that matters wouldnâ€™t it?

Fred

If I understand your proposition, if you deviated leftward, the left-controlling system would have error but the right-controller would not. So the left system would pull the car back toward the centre, while the right system would do nothing. If you deviated to the right, the opposite would be true. The two one-way systems work together like one two-way system. What matters is not the magnitude of the error signal, but the difference between the error signals in the two systems.

At least that’s the basic function. You can get a lot more out of such a system if you can bias the zero-error point where perception equals reference of each one-way system. If the left system’s zero error doesn’t happen until there is a little rightward deviation, and vice-versa, when the overall error is zero, the two systems oppose one another, like the two muscles that make a joint angle move one way or the other. With overlap, you get a “stiff” control region because the two systems are in Kent McClelland’s conflictive collective control state, where the apparent loop gain is the sum of the individual gains. If the “zero-error” bias is the other way, so there’s a region over which neither system produces an error signal, you get a tolerance zone.

I don’t know whether any of this actually happens in a nervous system, but as soon as you talk about opposed one-way control systems, these possibilities become things to be asked about.

Martin

From: Richard Marken [mailto:rsmarken@gmail.com]
Sent: Monday, August 14, 2017 3:46 PM
To: csgnet@lists.illinois.edu
Subject: Re: Control by Nervous System (was Re: Are academics stuck…)

[From Rick Marken (2017.08.14. 1245)]

Fred Nickols (2017.08.14.1426 ET)–

FN: Iâ€™m probably way out of my depth here but I donâ€™t know that the comparator function needs to assign polarity (i.e., negative or positive). What if all the comparator detects is the magnitude of the error? Then it wouldnâ€™t matter which was greater than the other.

RM: This is equivalent to saying that the comparator function is |r-p|, the absolute value of the difference between r and p. The problem with this is that, when p is greater than the reference, the output of the comparator must drive system output so as to reduce p in proportion to the degree to which it exceeds r; that is, the feedback effects of output on input must be negative. If the error signal produced by the comparator were proportional only to the magnitude of the difference between p and r (rather than to its direction – greater than or less than r – as well) , then the feedback would not be negative when p exceeded r and there would be no control in that range of values of p.

Best

Rick

Fred Nickols

From: Martin Taylor [mailto:mmt-csg@mmtaylor.net]
Sent: Monday, August 14, 2017 1:52 PM
To: csgnet@lists.illinois.edu
Subject: Re: Control by Nervous System (was Re: Are academics stuck…)

[Martin Taylor 2017.08.14.13.41]

[From Rick Marken (2017.08.14.0920)]

…

RM: There is one fundamental question about how control systems are implemented in the nervous system that, to my knowledge, has not been confirmed by neurophysiological observation. That is the question of how the nervous system handles the fact that control requires that error signals be both positive and negative.

RM: The PCT model of control by the nervous system assumes that control signals – perceptual, reference, error and output – are carried as neural firing rates, which can only be positive. So how does the nervous system implement the “comparator” function, r - p? If p < r then there is no problem; the output of this function, the error signal, is a positive neural firing rate. However, if p > r then the output of this function is negative and there is no such thing as a negative rate of neural firing.

RM: The obvious solution to this problem is to have control systems be made up of two parallel comparators that receive the same perceptual and reference signal inputs but do the comparison function in the opposite ways and produces separate output (error) signals; one comparator function is r-p and the other is p - r. When p < r the r-p comparator produces an error signal while the p-r comparator output remains at 0; when p > r, the p-r comparator produces an error signal while the r - p comparator remains at 0. The outputs of each of these comparators are always positive so they must be connected to the output function so that the output of the r-p comparator excites and the output of the p-r comparator inhibits output.

RM: I am not aware of any neurophysiological evidence for the existence of such a “parallel comparator” architecture that would support the existence of control systems in the nervous system. It seems like this is one of the most basic implications of the control model for the neurophysiological basis of perceptual control. So while the “parallel comparator” system does provide a model of how control systems “might operate within the nervous system”, it doesn’t prove that they do operate this way. Since it seems that there are some neurophysiologists on CSGNet I wonder if they know of any evidence for the existence of the kind of “parallel comparator” neural architecture required to implement control systems in the nervous system.

I second this motion! That issue has always been lurking in the background. The argument has always been from necessity, that since we can observe the behavioural manifestations of control from either side of an apparent reference value of a CEV in the environment, something in the brain must treat both directions of error more or less the same. Energetically, it would be very wasteful for zero error in every control unit to be signalled by a firing rate about midway between zero and the maximum possible, so the balance presumably isn’t done that way. It would be really nice to know the mechanism or mechanisms that allow this to happen. We have opposed musculature that does the physical job in some cases. We can see that. But is there anything equivalent in the nervous system?

Martin

–

Richard S. Marken

"Perfection is achieved not when you have nothing more to add, but when you
have nothing left to take away.â€?
–Antoine de Saint-Exupery

MartinT · August 15, 2017, 1:30pm

[From Fred
Nickols (2017.08.15.0631)]

Â

        Iâ€™m puzzled

by the thermostat example in your second paragraph, Martin,
the one referring to turning on the furnace or air
conditioner a little or a lot.

Â

        I thought

furnaces and air conditioners were simply on or off and the
hot air or cold air was essentially a fixed temperature.Â
Thus, the way they raise or lower the room temperature has
to do mainly with how long theyâ€™re on or off, not how cold
or hot their output is.Â Hence, your statement in the first
paragraph that all the thermostat can do is turn the furnace
or air conditioner on or off.

Â

        Am I missing

something?

···

From: Martin Taylor
[mailto:mmt-csg@mmtaylor.net]
Sent: Monday, August 14, 2017 11:46 PM
To: csgnet@lists.illinois.edu
Subject: Re: Control by Nervous System (was Re:
Are academics stuck…)

Â

[Martin Taylor 2017.08.14.23.33]

          [From Fred
Nickols (2017.08.14.1725 ET)]

Â

          Gee,
thanks, Martin.Â I guess Iâ€™m not so stupid after all.Â
Thanks for stating my proposition more clearly than I did.

Â

          One
question: Why does the magnitude of the error signal not
matter.Â If itâ€™s a big leftward deviation and it takes a
big change in steering to reduce that error, doesnâ€™t
magnitude matter?

      I didn't mean to imply that the error magnitude doesn't

matter. In the kinds of tracking systems we often simulate, it
does matter. In an ordinary thermostat, it doesn’t. All the
thermostat can do is turn the furnace or air conditioner on or
off. If it turns both on together, it wastes energy. If there
is a temperature range in which neither is on, the temperature
can vary freely over this range.

      Whether the magnitude of the error matters depends on what you

are trying to control and what you can do about it. You could
have a tracking system in which it doesn’t matter (give the
subject only a switch to move the cursor up, don’t move, or
down, and you could have a thermostat in which it does matter:
turn the furnace or air conditioner on a little or a lot
depending on how far the temperature is from its current set
point.

      Martin

          Fred

Nickols

Â

From: Martin Taylor [mailto:mmt-csg@mmtaylor.net ]
Sent: Monday, August 14, 2017 5:04 PM
To: csgnet@lists.illinois.edu
Subject: Re: Control by Nervous System (was Re:
Are academics stuck…)

Â

[Martin Taylor 2017.08.14.16.45]

          [From Fred Nickols (2017.08.14.1630

ET)]

Â

          If Iâ€™m trying to maintain my car in the
center of the lane, what if Iâ€™ve got one system that
controls leftward deviation and one that controls
rightward deviation?Â Then magnitude of the error signal
would be all that matters wouldnâ€™t it?

Â

Fred

        If I understand your proposition, if you deviated leftward,
the left-controlling system would have error but the
right-controller would not. So the left system would pull
the car back toward the centre, while the right system would
do nothing. If you deviated to the right, the opposite would
be true. The two one-way systems work together like one
two-way system. What matters is not the magnitude of the
error signal, but the difference between the error signals
in the two systems.
        At least that's the basic function. You can get a lot more
out of such a system if you can bias the zero-error point
where perception equals reference of each one-way system. If
the left system’s zero error doesn’t happen until there is a
little rightward deviation, and vice-versa, when the overall
error is zero, the two systems oppose one another, like the
two muscles that make a joint angle move one way or the
other. With overlap, you get a “stiff” control region
because the two systems are in Kent McClelland’s conflictive
collective control state, where the apparent loop gain is
the sum of the individual gains. If the “zero-error” bias is
the other way, so there’s a region over which neither system
produces an error signal, you get a tolerance zone.
        I don't know whether any of this actually happens in a
nervous system, but as soon as you talk about opposed
one-way control systems, these possibilities become things
to be asked about.
        Martin

Â

From: Richard Marken [mailto:rsmarken@gmail.com ]
Sent: Monday, August 14, 2017 3:46 PM
To: csgnet@lists.illinois.edu
Subject: Re: Control by Nervous System (was Re: Are
academics stuck…)

Â

                [>From Rick Marken

(2017.08.14. 1245)]

Â

                      Fred

Nickols (2017.08.14.1426 ET)–

Â

                      FN:
Iâ€™m probably way out of my depth here but I
donâ€™t know that the comparator function needs
to assign polarity (i.e., negative or
positive).Â What if all the comparator detects
is the magnitude of the error? Then it
wouldnâ€™t matter which was greater than the
other.

Â

                  RM: This is equivalent to
saying that the comparator function is |r-p|, the
absolute value of the difference between r and p.
The problem with this is that, when p is greater
than the reference, the output of the comparator
must drive system output so as to reduce p in
proportion to the degree to which it exceeds r;
that is, the feedback effects of output on input
must be negative. If the error signal produced by
the comparator were proportional only to the
magnitude of the difference between p and r
(rather than to its direction – greater than or
less than r – as well) , then the feedback would
not be negative when p exceeded r and there would
be no control in that range of values of p.Â

Â

BestÂ

Â

Rick

Â

                      Fred

Nickols

Â

From:
Martin Taylor [mailto:mmt-csg@mmtaylor.net ]
Sent: Monday, August 14, 2017 1:52
PM
To: csgnet@lists.illinois.edu
Subject: Re: Control by Nervous
System (was Re: Are academics stuck…)

Â

[Martin Taylor 2017.08.14.13.41]

                              [From

Rick Marken (2017.08.14.0920)]

...

                                RM:
There is one fundamental question
about how control systems are
implemented in the nervous system
that, to my knowledge, has not been
confirmed by neurophysiological
observation. That is the question of
how the nervous system handles the
fact that control requires that
error signals be both positive and
negative.Â

Â

                                RM:
The PCT model of control by the
nervous system assumes that control
signals – perceptual, reference,
error and output – are carried as
neural firing rates, which can only
be positive. So how does the nervous
system implement the “comparator”
function, r - p?Â If p < r then
there is no problem; the output of
this function, the error signal, is
a positive neural firing rate.
However, if p > r then the output
of this function is negative and
there is no such thing as a negative
rate of neural firing.Â

Â

                                RM:
The obvious solution to this problem
is to have control systems be made
up of two parallel comparators that
receive the same perceptual and
reference signal inputs but do the
comparison function in the opposite
ways and produces separate output
(error) signals; one comparator
function is r-p and the other is p -
r. When p < r the r-p comparator
produces an error signal while the
p-r comparator output remains at 0;
when p > r, the p-r comparator
produces an error signal while the r

p comparator remains at 0. The
outputs of each of these comparators
are always positive so they must be
connected to the output function so
that the output of the r-p
comparator excites and the output of
the p-r comparator inhibits output.Â

Â

                                RM:
I am not aware of any
neurophysiological evidence for the
existence of such a “parallel
comparator” architecture that would
support the existence of control
systems in the nervous system. It
seems like this is one of the most
basic implications of the control
model for the neurophysiological
basis of perceptual control. So
while Â the “parallel comparator”
system does provide a model of how
control systems “might operate
within the nervous system”, it
doesn’t prove that they do operate
this way. Since it seems that there
are some neurophysiologists on
CSGNet I wonder if they know of any
evidence for the existence of the
kind of “parallel comparator” neural
architecture required to implement
control systems in the nervous
system.Â

Â

                          I
second this motion! That issue has always
been lurking in the background. The
argument has always been from necessity,
that since we can observe the behavioural
manifestations of control from either side
of an apparent reference value of a CEV in
the environment, something in the brain
must treat both directions of error more
or less the same. Energetically, it would
be very wasteful for zero error in every
control unit to be signalled by a firing
rate about midway between zero and the
maximum possible, so the balance
presumably isn’t done that way. It would
be really nice to know the mechanism or
mechanisms that allow this to happen. We
have opposed musculature that does the
physical job in some cases. We can see
that. But is there anything equivalent in
the nervous system?
                          Martin

Â

–

                                      Richard S.

MarkenÂ

                                        "Perfection

is achieved not when you
have nothing more to add,
but when you

                                        have nothing left to take

away.â€?

                                        Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â

Â Â --Antoine de
Saint-Exupery

Â

Â

FredNickols · August 15, 2017, 2:01pm

[From Fred Nickols (2017.08.15.1001)]

Aha.Â I get it.Â Thanks.

Fred

···

From: Martin Taylor [mailto:mmt-csg@mmtaylor.net]
Sent: Tuesday, August 15, 2017 9:30 AM
To: csgnet@lists.illinois.edu
Subject: Re: Control by Nervous System (was Re: Are academics stuck…)

[Martin Taylor 2017.08.15.09.27]

[From Fred Nickols (2017.08.15.0631)]

Iâ€™m puzzled by the thermostat example in your second paragraph, Martin, the one referring to turning on the furnace or air conditioner a little or a lot.

I thought furnaces and air conditioners were simply on or off and the hot air or cold air was essentially a fixed temperature. Thus, the way they raise or lower the room temperature has to do mainly with how long theyâ€™re on or off, not how cold or hot their output is. Hence, your statement in the first paragraph that all the thermostat can do is turn the furnace or air conditioner on or off.

Am I missing something?

No, I am. I missed adding some word such as “unconventional” about the furnace and air conditioner. I was imagining ones that could deliver or take out heat at a variable rate, for example an electric furnace or an air-conditioner with a variable rate pump.

Martin

Fred Nickols

From: Martin Taylor [mailto:mmt-csg@mmtaylor.net]
Sent: Monday, August 14, 2017 11:46 PM
To: csgnet@lists.illinois.edu
Subject: Re: Control by Nervous System (was Re: Are academics stuck…)

[Martin Taylor 2017.08.14.23.33]

[From Fred Nickols (2017.08.14.1725 ET)]

Gee, thanks, Martin. I guess Iâ€™m not so stupid after all. Thanks for stating my proposition more clearly than I did.

One question: Why does the magnitude of the error signal not matter. If itâ€™s a big leftward deviation and it takes a big change in steering to reduce that error, doesnâ€™t magnitude matter?

I didn’t mean to imply that the error magnitude doesn’t matter. In the kinds of tracking systems we often simulate, it does matter. In an ordinary thermostat, it doesn’t. All the thermostat can do is turn the furnace or air conditioner on or off. If it turns both on together, it wastes energy. If there is a temperature range in which neither is on, the temperature can vary freely over this range.

Whether the magnitude of the error matters depends on what you are trying to control and what you can do about it. You could have a tracking system in which it doesn’t matter (give the subject only a switch to move the cursor up, don’t move, or down, and you could have a thermostat in which it does matter: turn the furnace or air conditioner on a little or a lot depending on how far the temperature is from its current set point.

Martin

Fred Nickols

From: Martin Taylor [mailto:mmt-csg@mmtaylor.net]
Sent: Monday, August 14, 2017 5:04 PM
To: csgnet@lists.illinois.edu
Subject: Re: Control by Nervous System (was Re: Are academics stuck…)

[Martin Taylor 2017.08.14.16.45]

[From Fred Nickols (2017.08.14.1630 ET)]

If Iâ€™m trying to maintain my car in the center of the lane, what if Iâ€™ve got one system that controls leftward deviation and one that controls rightward deviation? Then magnitude of the error signal would be all that matters wouldnâ€™t it?

Fred

If I understand your proposition, if you deviated leftward, the left-controlling system would have error but the right-controller would not. So the left system would pull the car back toward the centre, while the right system would do nothing. If you deviated to the right, the opposite would be true. The two one-way systems work together like one two-way system. What matters is not the magnitude of the error signal, but the difference between the error signals in the two systems.

At least that’s the basic function. You can get a lot more out of such a system if you can bias the zero-error point where perception equals reference of each one-way system. If the left system’s zero error doesn’t happen until there is a little rightward deviation, and vice-versa, when the overall error is zero, the two systems oppose one another, like the two muscles that make a joint angle move one way or the other. With overlap, you get a “stiff” control region because the two systems are in Kent McClelland’s conflictive collective control state, where the apparent loop gain is the sum of the individual gains. If the “zero-error” bias is the other way, so there’s a region over which neither system produces an error signal, you get a tolerance zone.

I don’t know whether any of this actually happens in a nervous system, but as soon as you talk about opposed one-way control systems, these possibilities become things to be asked about.

Martin

From: Richard Marken [mailto:rsmarken@gmail.com]
Sent: Monday, August 14, 2017 3:46 PM
To: csgnet@lists.illinois.edu
Subject: Re: Control by Nervous System (was Re: Are academics stuck…)

[>From Rick Marken (2017.08.14. 1245)]

Fred Nickols (2017.08.14.1426 ET)–

FN: Iâ€™m probably way out of my depth here but I donâ€™t know that the comparator function needs to assign polarity (i.e., negative or positive). What if all the comparator detects is the magnitude of the error? Then it wouldnâ€™t matter which was greater than the other.

RM: This is equivalent to saying that the comparator function is |r-p|, the absolute value of the difference between r and p. The problem with this is that, when p is greater than the reference, the output of the comparator must drive system output so as to reduce p in proportion to the degree to which it exceeds r; that is, the feedback effects of output on input must be negative. If the error signal produced by the comparator were proportional only to the magnitude of the difference between p and r (rather than to its direction – greater than or less than r – as well) , then the feedback would not be negative when p exceeded r and there would be no control in that range of values of p.

Best

Rick

Fred Nickols

From: Martin Taylor [mailto:mmt-csg@mmtaylor.net]
Sent: Monday, August 14, 2017 1:52 PM
To: csgnet@lists.illinois.edu
Subject: Re: Control by Nervous System (was Re: Are academics stuck…)

[Martin Taylor 2017.08.14.13.41]

[From Rick Marken (2017.08.14.0920)]

…

RM: There is one fundamental question about how control systems are implemented in the nervous system that, to my knowledge, has not been confirmed by neurophysiological observation. That is the question of how the nervous system handles the fact that control requires that error signals be both positive and negative.

RM: The PCT model of control by the nervous system assumes that control signals – perceptual, reference, error and output – are carried as neural firing rates, which can only be positive. So how does the nervous system implement the “comparator” function, r - p? If p < r then there is no problem; the output of this function, the error signal, is a positive neural firing rate. However, if p > r then the output of this function is negative and there is no such thing as a negative rate of neural firing.

RM: The obvious solution to this problem is to have control systems be made up of two parallel comparators that receive the same perceptual and reference signal inputs but do the comparison function in the opposite ways and produces separate output (error) signals; one comparator function is r-p and the other is p - r. When p < r the r-p comparator produces an error signal while the p-r comparator output remains at 0; when p > r, the p-r comparator produces an error signal while the r - p comparator remains at 0. The outputs of each of these comparators are always positive so they must be connected to the output function so that the output of the r-p comparator excites and the output of the p-r comparator inhibits output.

RM: I am not aware of any neurophysiological evidence for the existence of such a “parallel comparator” architecture that would support the existence of control systems in the nervous system. It seems like this is one of the most basic implications of the control model for the neurophysiological basis of perceptual control. So while the “parallel comparator” system does provide a model of how control systems “might operate within the nervous system”, it doesn’t prove that they do operate this way. Since it seems that there are some neurophysiologists on CSGNet I wonder if they know of any evidence for the existence of the kind of “parallel comparator” neural architecture required to implement control systems in the nervous system.

I second this motion! That issue has always been lurking in the background. The argument has always been from necessity, that since we can observe the behavioural manifestations of control from either side of an apparent reference value of a CEV in the environment, something in the brain must treat both directions of error more or less the same. Energetically, it would be very wasteful for zero error in every control unit to be signalled by a firing rate about midway between zero and the maximum possible, so the balance presumably isn’t done that way. It would be really nice to know the mechanism or mechanisms that allow this to happen. We have opposed musculature that does the physical job in some cases. We can see that. But is there anything equivalent in the nervous system?

Martin

–

Richard S. Marken

"Perfection is achieved not when you have nothing more to add, but when you
have nothing left to take away.â€?
–Antoine de Saint-Exupery

rsmarken · August 15, 2017, 7:12pm

[From Rick Marken (2017.08.15.1212)]

···

Martin Taylor (2017.08.14.13.41)–

MT: I second this motion! That issue has always been lurking in the
background. The argument has always been from necessity, that since
we can observe the behavioural manifestations of control from either
side of an apparent reference value of a CEV in the environment,
something in the brain must treat both directions of error more or
less the same.

Â

RM: Maybe this isn’t as big a problem as I think it is. At the lowest levels of control it’s apparently dealt with using muscle groups that generate forces in opposite directions. You still need two parallel controllers getting the same reference signal with comparators computing the error in opposite ways (r-p and p-r). I think such circuits have been found. Maybe the neuro- physiologists in the group can confirm or deny this. Neural signals at higher levels seem like they may be more of a problem. But maybe not. I think our resident robot experts could help out on this one. How do roboticists deal with the “polarity” of control system output, particularly at higher levels?

      RM: There is one fundamental question about how control

systems are implemented in the nervous system that, to my
knowledge, has not been confirmed by neurophysiological
observation. That is the question of how the nervous system
handles the fact that control requires that error signals be
both positive and negative.Â

Best

Rick

–
Richard S. MarkenÂ

"Perfection is achieved not when you have nothing more to add, but when you
have nothing left to take away.â?
Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â --Antoine de Saint-Exupery

FredNickols · August 15, 2017, 7:46pm

[From Fred Nickols (2017.08.15.1545 ET)]

I just posted a diagram that elaborates on my earlier explanation of how the limitation of positive neural signals might be dealt with.

Fred Nickols

···

From: Richard Marken [mailto:rsmarken@gmail.com]
Sent: Tuesday, August 15, 2017 3:13 PM
To: csgnet@lists.illinois.edu
Subject: Re: Control by Nervous System (was Re: Are academics stuck…)

[From Rick Marken (2017.08.15.1212)]

Martin Taylor (2017.08.14.13.41)–

RM: There is one fundamental question about how control systems are implemented in the nervous system that, to my knowledge, has not been confirmed by neurophysiological observation. That is the question of how the nervous system handles the fact that control requires that error signals be both positive and negative.

MT: I second this motion! That issue has always been lurking in the background. The argument has always been from necessity, that since we can observe the behavioural manifestations of control from either side of an apparent reference value of a CEV in the environment, something in the brain must treat both directions of error more or less the same.

RM: Maybe this isn’t as big a problem as I think it is. At the lowest levels of control it’s apparently dealt with using muscle groups that generate forces in opposite directions. You still need two parallel controllers getting the same reference signal with comparators computing the error in opposite ways (r-p and p-r). I think such circuits have been found. Maybe the neuro- physiologists in the group can confirm or deny this. Neural signals at higher levels seem like they may be more of a problem. But maybe not. I think our resident robot experts could help out on this one. How do roboticists deal with the “polarity” of control system output, particularly at higher levels?

Best

Rick

–

Richard S. Marken

"Perfection is achieved not when you have nothing more to add, but when you
have nothing left to take away.â€?
–Antoine de Saint-Exupery

_Boris_Hartman · August 19, 2017, 11:08pm

Sorry Rupert to join discussion…

···

From: Richard Marken [mailto:rsmarken@gmail.com]
Sent: Monday, August 14, 2017 6:19 PM
To: csgnet@lists.illinois.edu
Cc: Richard Marken
Subject: Control by Nervous System (was Re: Are academics stuck…)

[From Rick Marken (2017.08.14.0920)]

Rupert Young (2017.08.13 17.00)–

RY: I refer you to the paper on p170 of this large document, “Simulation of Human Balance Control Using an Inverted,Pendulum Model,” and offer for your delectation (and despair) the comments in the abstract,

“It is probable that there are many concurrent control loops occurring in the central nervous system” and,
“no models of how these controllers might operate within the nervous system have yet been developed.”

RY: Below is my email to the authors…

RY: Actually there is a very detailed model of the functional control architecture of the nervous system, which you can see in the book “Behavior: The Control Of Perception”

RM: There is one fundamental question about how control systems are implemented in the nervous system that, to my knowledge, has not been confirmed by neurophysiological observation.

HB : Right… according to your knowledge ? Is yyour knowledge something special what should have some public value ?

RM : That is the question of how the nervous system handles the fact that control requires that error signals be both positive and negative.

HB : Well Rick. I think that beside Bills’ books you should read some others too…

/div>

RM: The PCT model of control by the nervous system assumes that control signals – perceptual, reference, error and output – are carried as neural firing rates, which can only be positive.

HB : I don’t know what this has to do with the events in nervous system ? Where did you get this idea ? Where did you see this in PCT literature ?Â In which sense positive ? The point of Â»transductionÂ« of Â»chemical substancesÂ« is somewhere else. Rick you should read more books before you came with such an obscure knowledge into public…

Bill P (B:CP) :

Therefore in the summation of neural currents, some currents contribute positivelly to the net excitation of the receiving cell, while others contribute negativelly.

HB :

Events in nervous system are much more complicated than you can imagine Rick. Why do you think sciences like physiology and neurophysiology developed ?

Bill P (B:CP) : Â

In a typical analysis in a nerve network is described in terms of time states

HB : Bill studied also a lot of other literature Rick. He was very educated man. I don’t take you as educated man and I think you knowledge is 1/100 of Bills.

You Rick wrongly focused on Â»firing ratesÂ« of nerv cells to understand what is happening in nervous system.

RM : So how does the nervous system implement the “comparator” function, r - p? If p < r then there is no problem; the output of this function, the error signal, is a positive neural firing rate. However, if p > r then the output of this function is negative and there is no such thing as a negative rate of neural firing.

HB : Rick as I said before. Start reading books…

RM: The obvious solution to this problem is to have control systems be made up of two parallel comparators that receive the same perceptual and reference signal inputs but do the comparison function in the opposite ways and produces separate output (error) signals;

HB : Do you understand Rick that you are talking about neurons ? All neurons have the same characteristics. They operate on the same principles. And you don’t know them, but still you are confussing all arround.

RM : ….one comparator function is r-p and the other iss p - r.

HB : What an imagination ? Start reading books Rick….

HB : When p < r the r-p comparator produces an error signal while the p-r comparator output remains at 0; when p > r, the p-r comparator produces an error signal while the r - p comparator remains at 0. The outputs of each of these comparators are always positive so they must be connected to the output function so that the output of the r-p comparator excites and the output of the p-r comparator inhibits output.

HB : Again your imagination at work and of course multilevel confussion on CSGnet…

/p>

RM: I am not aware of any neurophysiological evidence for the existence of such a “parallel comparator” architecture that would support the existence of control systems in the nervous system.

HB : So you admitt that you are working without any trace of evidence that such a construct as Â»two different comparatorsÂ« could exist, but still you are confussing CSGnet ???

RM : It seems like this is one of the most basic implications of the control model for the neurophysiological basis of perceptual control.

HB : You don’t know anything about basic mechnisms in nerv net and still you are concluding on some basic implications to the control model… You just said it…

RM: I am not aware of any neurophysiological evidence for the existence of such a “parallel comparator” architecture…

HB : You are not aware because your little knowledege doesn’ allow you to be aware of how nervous system function… Maybe you shoould read again Henry Yin or ask him what you don’t understand about nervous system, before you came out with such an confussing idea.

RM : So while the “parallel comparator” system does provide a model of how control systems “might operate within the nervous system”, it doesn’t prove that they do operate this way.

HB : You are admitting again that you have no proofs, no knowledge about nervous system and still you are concluding how might nervous system operate ???

RM : Since it seems that there are some neurophysiologists on CSGNet I wonder if they know of any evidence for the existence of the kind of “parallel comparator” neural architecture required to implement control systems in the nervous system.

HB : Some of this sort of mechanisms you try to guess are already in Bills’ literature… Start reading Rick, stop confussingg…

Boris

Best

Rick

–

Richard S. Marken

"Perfection is achieved not when you have nothing more to add, but when you
have nothing left to take away.â€?
–Antoine de Saint-Exupery

_Boris_Hartman · August 20, 2017, 12:07am

Sorry Martin to jump in…

···

From: Richard Marken [mailto:rsmarken@gmail.com]
Sent: Tuesday, August 15, 2017 9:13 PM
To: csgnet@lists.illinois.edu
Subject: Re: Control by Nervous System (was Re: Are academics stuck…)

[From Rick Marken (2017.08.15.1212)]

Martin Taylor (2017.08.14.13.41)–

RM: There is one fundamental question about how control systems are implemented in the nervous system that, to my knowledge, has not been confirmed by neurophysiological observation. That is the question of how the nervous system handles the fact that control requires that error signals be both positive and negative.

MT: I second this motion! That issue has always been lurking in the background. The argument has always been from necessity, that since we can observe the behavioural manifestations of control from either side of an apparent reference value of a CEV in the environment, something in the brain must treat both directions of error more or less the same.

HB : I also think so, that Â»the brain must treat both directions of error more or less the sameÂ«… since the Â»comparatorsÂ« work the same…
:p>

RM: Maybe this isn’t as big a problem as I think it is. At the lowest levels of control it’s apparently dealt with using muscle groups that generate forces in opposite directions.

HB : It’s not about generating forces in opposite direction it’s about the same mechanisms that produce Â»generating forces in opposite directionsÂ«….

<
RM : You still need two parallel controllers getting the same reference signal with comparators computing the error in opposite ways (r-p and p-r).

HB : You just take any physiological book and look how this is functioning so how Â»opposite forcesÂ« are produced…

RM : I think such circuits have been found.

HB : You think or you know or you are guessing ?

RM earlier :

I am not aware of any neurophysiological evidence for the existence of such a “parallel comparator” architecture that would support the existence of control systems in the nervous system.

RM : Maybe the neuro- physiologists in the group can confirm or deny this. Neural signals at higher levels seem like they may be more of a problem. But maybe not. I think our resident robot experts could help out on this one. How do roboticists deal with the “polarity” of control system output, particularly at higher levels?

HB : Well I always think that imagination has no limits…. But as I said before, you can start reading Bills’ books and you’ll find many answers to these questionsâ€¦

Boris

Best

Rick

–

Richard S. Marken

"Perfection is achieved not when you have nothing more to add, but when you
have nothing left to take away.â€?
–Antoine de Saint-Exupery