Adaptive Resonance Theory

I was just watching this YouTube video

and I thought of HTM. Has resonance theory been considered? If so, what implications might that hold for HTM? Here are some relevant links/papers:

https://arxiv.org/abs/1905.11437

https://www.pbs.org/wgbh/nova/article/the-good-vibrations-of-quantum-field-theories/

https://theconversation.com/profiles/tam-hunt-517188/articles

https://theconversation.com/could-consciousness-all-come-down-to-the-way-things-vibrate-103070

http://www.scholarpedia.org/article/Self-organization_of_brain_function

https://theconversation.com/how-can-you-tell-if-another-person-animal-or-thing-is-conscious-try-these-3-tests-115835

3 Likes

Personally, I haven’t seen any need to consider it. Explaining the sensation of consciousness is like explaining why a bell doesn’t sound like the color red. What’s more interesting to me is to explain what function it serves in a brain, and does that function really require resonating quantum states?

6 Likes

Also found it. Looks like a huge theory. It’s not only Adaptive Resonance Theory, but also Complementary Computing theory

1 Like

Resonators can be:
1/ Over damped.
2/ Critically damped.
3/ Under damped.
Which can be measured as a Q value.

Usually some positive feedback is needed for oscillation:
1/ A small amount of positive feedback will just boost the Q value. (Regeneration)
2/ A large amount of positive feedback will set the system oscillating.
3/ In some cases positive feedback can cause hysteresis where the system locks in one state or another rather than oscillate.

If the gain of the thing providing the positive feedback increases with amplitude then the amplitude of oscillation will increase until the system slams into some physical constraint, voltage, energy limitations, whatever.
If the gain is exactly constant with amplitude (difficult in real physical systems)
then an oscillation will continue with whatever amplitude it started with.
If the gain decreases with amplitude the the system will oscillate at some moderate amplitude where the loop gain around the system becomes exactly 1.

And it can get even more complicated than that when there are other temporal aspects involved.

1 Like

I have internalized much of Grossberg’s ideas on his ART models. I see that as a good approximation of how HTM responds to inputs - resonating with known inputs. HTM adds that it resonates with known sequences.

As far as comparison to the traditional resonance theory - a good starting point is to think of it as a piano where the closest matching string starts to sing out as long as the input sound is present.

That model only gets you so far in understanding as the piano is a fully parallel system, where neural networks mix the various internal representations together.

So - how is it different? When you apply the input it tries to respond but perhaps the resonance is poor as some stronger learned pattern is trying to respond even though it really is a poor match.

A key feature is that patterns that are not a good match get tired and fade out, allowing the weaker but better match patterns to assert themselves and resonate. Habituation.

This is the adaptive part of the ART name.

Naturally, there is a learning model that goes with all this.

This is the very first time I encountered a system that searches through its internal representations to find a good match.

You may find his instar and outstar neurons to be strange but on reflection - it does capture the key biological features necessary for his model.

Even if you are fully invested in HTM theory there is much to be learned reading Grossberg’s works - It is like learning a new language - it may help you better understand your mother tongue.

He has been doing this stuff forever and really knows his neuroscience.

If you are interested here are some links to check out:


http://www.brainm.com/software/pubs/brain/Grossberg-adaptive-resonance.pdf


3 Likes

Thanks, @bitking. Joe in the video did such a poor job explaining ART (bringing up panpsychism, proto consciousness, and quantum resonance fields) that I didn’t actually take the time to look into it any further. I see that the theory is not so “woo woo” after all. Definitely looks worth digging into further.

4 Likes

“Harmonic resonance is an extraordinarily diverse and varied phenomenon seen in countless forms throughout the universe, from gravitational orbital resonances, to electromagnetic oscillations, to acoustical vibrations in solids, liquids, and gases, to laser resonance in light and microwaves. Harmonic resonance spans a vast range of spatial scales, from the tiniest wave-like vibrations of the elemental particles of matter, to orbital resonances that emerge from spinning disks of gas and stars. But across this vast range of spatial scales and diverse media, there are certain general properties of harmonic resonance that are common to all of them. They all tend to oscillate at some characteristic frequency, and at its higher harmonics, frequencies that are integer multiples of the fundamental frequency. They all exhibit spatial standing waves, whose wavelength is inversely proportional to their frequencies. They all tend to subdivide one, two, or three-dimensional spaces into equal intervals of alternating reciprocating forces dynamically balanced against each other, with the twin properties of periodicity and symmetry across every possible dimension of space and time. These and many other properties, are properties of resonance in the abstract, manifested across all those diverse forms and media. Harmonic resonance is a higher order organizational principle of physical matter, that transcends any particular implementation in a physical medium. It is the properties of that transcendent, more general concept of harmonic resonance that are the focus of this book, because it is those transcendant properties that reveal the essential properties of resonance itself, and explain how those properties lead to the emergence of mind from brain.”
http://cns-alumni.bu.edu/~slehar/HRezBook/Chap2.pdf

Currently considered to be primarily responsible for the production and circulation of cerebral spinal fluid, the ventricle system may appear an overly elaborate structure for a bath cushion, unless perhaps like the inner ear in detecting motion and sound, such fluid-filled chambers serve an important role in perception by means of mechanosensation such as efficient transduction and amplification of energy waves in a medium. (Our model predicted the possibility of resonant chambers in the brain to which we observed the ventricular system (VS), and so considered what role such an elaborate structure might play in perception or recognition and looked to the inner ear as example.)

Beautiful depiction of the Ventricular System:
https://commons.wikimedia.org/wiki/File:Human_Ventricular_system_colored_and_animated.gif#/media/File:Human_Ventricular_system_colored_and_animated.gif

“The ventricular system is a set of four interconnected cavities (ventricles) in the brain, where the cerebrospinal fluid (CSF) is produced… allowing for the flow of CSF to circulate (and)… which bathes and cushions the brain and spinal cord within their bony confines”