Hi everyone,
I am wondering if any of you know whether or not anyone has tried to build a hierarchical model, such as Bill’s arm control model (or even a single-level model) using neural nets / spikes, etc, to underlie the control systems?
Cheers,
Heather
Canadian Centre for Behavioural Neuroscience
Behaviour and Evolution Research Group
Office EP1283A
University of Lethbridge
Canada
403.634.7075
Hi Heather, that’s an interesting question. I would imagine that Bill’s code for the arm demo is more complex than some apparent ‘neural networks’ in the literature - they don’t actually model the physical properties of neurones do they? The two by fourteen network of control
systems in the demo would be modelling some of the proposed functional properties of neural systems according to PCT. I can see that Bill made some educated guesses as to the micro-anatomical substrates of control systems in B:CP but I don’t know of any attempts to model these on a physical/chemical level.
Warren
Sent from my iPhone
On 18 Dec 2013, at 18:51, Heather Bell heather.bell@ULETH.CA wrote:
Hi everyone,
I am wondering if any of you know whether or not anyone has tried to build a hierarchical model, such as Bill’s arm control model (or even a single-level model) using neural nets / spikes, etc, to underlie the control systems?
Cheers,
Heather
Canadian Centre for Behavioural Neuroscience
Behaviour and Evolution Research Group
Office EP1283A
University of Lethbridge
Canada403.634.7075
[From Rick Marken (2013.12.19.0900)]
On Wed, Dec 18, 2013 at 11:55 PM, Warren Mansell wmansell@gmail.com wrote:
Hi Heather, that’s an interesting question. I would imagine that Bill’s code for the arm demo is more complex than some apparent ‘neural networks’ in the literature - they don’t actually model the physical properties of neurones do they? The two by fourteen network of control
systems in the demo would be modelling some of the proposed functional properties of neural systems according to PCT. I can see that Bill made some educated guesses as to the micro-anatomical substrates of control systems in B:CP but I don’t know of any attempts to model these on a physical/chemical level.
Warren
RM: These are great points, Warren. I would suggest, however, that Heather might find higher fidelity models of control by looking at robotic controllers, like the ones recently described by Rupert Young. The sensors in these models, to the extent that they derive perceptual signals based on electronics rather than program code, could be considered models of the neural nets that produce perceptions in living systems. I think it would be great to try to build working models of the neural nets that produce controllable perceptions. But that is a somewhat separate endeavor – though a very interesting and important one - from simply assuming that such networks exists and building models that include the functional; properties of these networks without specifying how these functions are carried out. For example, I’d love to see a model that includes the neural networks that produce a perceptual signals that represent the vertical and optical optical velocities that are presumably controlled when intercepting objects. But I don’t really need to see that such networks in order to build successful models of object interception that control these perceptions.
Best
Rick
Sent from my iPhone
On 18 Dec 2013, at 18:51, Heather Bell heather.bell@ULETH.CA wrote:
Hi everyone,
I am wondering if any of you know whether or not anyone has tried to build a hierarchical model, such as Bill’s arm control model (or even a single-level model) using neural nets / spikes, etc, to underlie the control systems?
Cheers,
Heather
Canadian Centre for Behavioural Neuroscience
Behaviour and Evolution Research Group
Office EP1283A
University of Lethbridge
Canada403.634.7075
–
Richard S. Marken PhD
www.mindreadings.com
The only thing that will redeem mankind is cooperation.
– Bertrand Russell
Hmmmm… I guess it wasn’t so much an application question… it seems like engineered control systems perform pretty well to do a wide variety of things, especially when there aren’t too many levels. But those aren’t built the same way they are in organisms. It was more of a question of what constitutes a control system that in an organism, and can we make one using similar components? Obviously, neurons are not the only component of at least low level controllers… but perhaps for higher level ones? And I agree that many network models aren’t very good (they’re not parallel, and/or they are digital, and/or they oversimplify neural function) but maybe building a hierarchical control system using something like a smallish network instantiated on a VLSI chip (analog) might be fun to try… once anyone gets one that’s more than 4 neurons to work consistently, I guess 
Anyway, it was just something I have been wondering about…
On Wed, Dec 18, 2013 at 11:55 PM, Warren Mansell wmansell@gmail.com wrote:
Hi Heather, that’s an interesting question. I would imagine that Bill’s code for the arm demo is more complex than some apparent ‘neural networks’ in the literature - they don’t actually model the physical properties of neurones do they? The two by fourteen network of control
systems in the demo would be modelling some of the proposed functional properties of neural systems according to PCT. I can see that Bill made some educated guesses as to the micro-anatomical substrates of control systems in B:CP but I don’t know of any attempts to model these on a physical/chemical level.
Warren
RM: These are great points, Warren. I would suggest, however, that Heather might find higher fidelity models of control by looking at robotic controllers, like the ones recently described by Rupert Young. The sensors in these models, to the extent that they derive perceptual signals based on electronics rather than program code, could be considered models of the neural nets that produce perceptions in living systems. I think it would be great to try to build working models of the neural nets that produce controllable perceptions. But that is a somewhat separate endeavor – though a very interesting and important one - from simply assuming that such networks exists and building models that include the functional; properties of these networks without specifying how these functions are carried out. For example, I’d love to see a model that includes the neural networks that produce a perceptual signals that represent the vertical and optical optical velocities that are presumably controlled when intercepting objects. But I don’t really need to see that such networks in order to build successful models of object interception that control these perceptions.
Best
Rick
Sent from my iPhone
On 18 Dec 2013, at 18:51, Heather Bell heather.bell@ULETH.CA wrote:
Hi everyone,
I am wondering if any of you know whether or not anyone has tried to build a hierarchical model, such as Bill’s arm control model (or even a single-level model) using neural nets / spikes, etc, to underlie the control systems?
Cheers,
Heather
Canadian Centre for Behavioural Neuroscience
Behaviour and Evolution Research Group
Office EP1283A
University of Lethbridge
Canada403.634.7075
–
Richard S. Marken PhD
www.mindreadings.com
The only thing that will redeem mankind is cooperation.
– Bertrand Russell
[David Goldstein (2013.12.20.04:03]
Heather,
I remember that Bill gave a talk at the U. of Colorado at Boulder to a group of psychologists. There were some neuroscience types that were interested in his work. Perhaps Barbara and Allie, his daughters, might know who he was talking to. Also, Dr. Henry Yin is a neuroscientist at Duke, who is part of part of the CSG. He might have some information and might be a worthwhile person to contact.
There is a book by Douglas S. Riggs which might be worth looking into.
David
Hmmmm… I guess it wasn’t so much an application question… it
seems like engineered control systems perform pretty well to do a wide variety of things, especially when there aren’t too many levels. But those aren’t built the same way they are in organisms. It was more of a question of what constitutes a control system that in an organism, and can we make one using similar components? Obviously, neurons are not the only component of at least low level controllers… but perhaps for higher
level ones? And I agree that many network models aren’t very good (they’re not parallel, and/or they are digital, and/or they oversimplify neural function) but maybe building a hierarchical control system using something like a smallish network instantiated on a VLSI chip (analog) might be fun to try… once anyone gets one that’s more than 4 neurons to work consistently, I guess
Anyway, it was just something I have been wondering about…
On Thursday, December 19, 2013 7:49 PM, Heather Bell heather.bell@ULETH.CA wrote:
On Dec 19, 2013 10:03 AM, “Richard Marken” rsmarken@gmail.com wrote:
[From Rick Marken (2013.12.19.0900)]
On Wed, Dec 18, 2013 at 11:55 PM, Warren Mansell wmansell@gmail.com wrote:
Hi Heather, that’s an interesting question. I would imagine that Bill’s code for the arm demo is more complex than some apparent ‘neural networks’ in the literature - they don’t actually model the physical properties of neurones do they? The two by fourteen network of control
systems in the demo would be modelling some of the proposed functional properties of neural systems according to PCT. I can see that Bill made some educated guesses as to the micro-anatomical substrates of control systems in B:CP but I don’t know of any attempts to model these on a physical/chemical level.
Warren
RM: These are great points, Warren. I would suggest, however, that Heather might find higher fidelity models of control by looking at robotic controllers, like the ones recently described by Rupert Young. The sensors in these models, to the extent that they derive perceptual signals based on electronics rather than program code, could be considered models of the neural nets that produce perceptions in living systems. I think it would be great to try to build working models of the neural nets that produce controllable perceptions. But that is a somewhat separate endeavor – though a very interesting and important one - from simply assuming that such networks exists and building models that include the functional; properties of these networks without specifying how these functions are carried out. For example, I’d love to see a model that includes the neural networks that produce a
perceptual signals that represent the vertical and optical optical velocities that are presumably controlled when intercepting objects. But I don’t really need to see that such networks in order to build successful models of object interception that control these perceptions.
Best
Rick
Sent from my iPhone
On 18 Dec 2013, at 18:51, Heather Bell heather.bell@ULETH.CA wrote:
Hi everyone,
I am wondering if any of you know whether or not anyone has tried to build a hierarchical model, such as Bill’s arm control model (or even a single-level model) using neural nets / spikes, etc, to underlie the control systems?
Cheers,
Heather
Canadian Centre for Behavioural Neuroscience
Behaviour and Evolution Research Group
Office EP1283A
University of Lethbridge
Canada403.634.7075
–
Richard S. Marken PhD
www.mindreadings.com
The only thing that will redeem mankind is cooperation.
– Bertrand Russell