Brain Physiology, Cognition and Consciousness group: topic
This is a public discussion board
Pattern Recognition Principle for a Theory of Mind
Gilberto de Paiva
Tuesday, 28 July 2009 01:19 UTC
Hello all,
I just finished and submited an article that proposes a theory of mind I developed based on the well known pattern recognition mechanism.
As this is a new idea in this forum I created this new topic.
The abstract is bellow:
“I propose that pattern recognition, memorization and processing are key concepts that can be a principle set for the theoretical modeling of the mind function. Most of the questions about the mind functioning can be answered by a descriptive modeling and definitions from these principles. An understandable consciousness definition can be drawn based on the assumption that a pattern recognition system can recognize its own patterns of activity. The principles, descriptive modeling and definitions can be a basis for theoretical and applied research on cognitive sciences, particularly at artificial intelligence studies.”
Link: http://arxiv.org/abs/0907.4509
Briefly, I’m proposing that the PATTERN RECOGNITION MECHANISM can be as important to understand the mind functionig as is the NATURAL SELECTION MECHANISM to understand the theory of evolution.
For me it is totaly clear that the mind is a pattern recognition and processing system as I describe in the article. I can understand and explain almost all the human mind, and also all the phylosophycal questions I was faced up with this theory. I would like to know your opinions about I’m proposing.
Thank you all, and I apologize for some gramatical errors I found later (I’ll try to fix them).
Gilberto de Paiva
Updated 28 July 2009 02:55 UTC
-
Replies
Jump to resultsResults
-
Pattern Recognitino is important, but first you have to have a representation of the pattern that is cmoplete enough that it can be recognized. How do you see this representation occuring in your theoretical system, or is it just assumed?
-
The basic patterns are already identified Gilberto. Recursion at the neurological level allows for creation of basic patterns and covers, using association with, for example, the visual system, the dynamics of the parafovea. Details analysis emerges from the ‘middle’ of such in the form of the fovea and the overall emphasis is on integrating (pattern matching, SAMENESS, symmetry, form detection) vs differentiating (DIFFERENCE, parts, anti-symmetry).
These same patterns apply to ALL sensory systems with local contexts adding refinements (or ignoring aspects).
The CONTEXT bias present works top-down and so regulates the set of possible degrees of freedom into a pattern. Thus the quality of WHOLENESS maps to a set of relationships that can resonate with some context to differentiate ‘whole’ from ‘part’. Recursion in the form of feedback then allows for finer distinctions etc (our drive to habituate is a drive to create patterns in the form of instincts/habits where such is energy-conserving).
The basic neurology covering differentiating/integrating elicits patterns grounded in a sense of wholeness (blending), partness (bounding), static relatedness (bonding), dynamic relatedness (binding), and their composites. LABELS then tie these ‘universals’ to local contexts and so emerges specialist languages that allow for communication out of the single context that is our basic nature as a neuron-dependent species.
We can map these basics to such as the classes of numbers we use to represent reality through mathematics and so, by association, all that is mathematical. We can also map these basics to classes of emotions and so our first ‘language’ operating within the immediate, local, context that is then extended through use of labels to communicate out of context.
The dynamics of meaning generation cover development from a ‘language of the vague’ that presents core patterns of meaning derived from neural activity utilising the recursion of the differentiate/integrate dichotomy; all covered in the IDM work where we focus on the use of analogy and metaphor in deriving many meanings from the one set of meanings grounded in the neurology (where analogy/metaphor are the core manifestations of pattern matching)
Chris.
-
BTW Gilberto, note that the IDM work covers recursion and includes the identification of an organic format that emerges from the original mechanistic format of recursion. In other words there is a braking system to ‘mindless recursion’ where any potential infinite regress is avoided/limited in that the organic nature covers the emergence of language to describe ‘all there is’. With this property of recursion comes such as ‘all is connected’ etc where we can use the language as a whole to describe all parts – it is, as such, self-referencing.
Thus ANY concept can be analysed to a level where a language forms allowing the concept to appear as able to describe itself – and that includes “Consciousness” ;-) – As such we can derive all of the POSSIBLE classes of consciousness where individual consciousness is a local context instance of a class.
Chris.
-
Graeme Smith Wrote:
“Pattern Recognitino is important, but first you have to have a representation of the pattern that is cmoplete enough that it can be recognized. How do you see this representation occuring in your theoretical system, or is it just assumed?”Graeme Smith,
Actually the pattern recognition concept is widelly studied in 2 ways:
- neural-biological (animal) systems of pattern recognition
- artifical (neural network, algoritm) generaly computer powered systems of pattern recognitionWhat I can interpret from the studies at animal brain recognition capabilities (also from the common sense) is that THE NEURAL SYSTEM HAVE THE CAPABILITY OF RECOGNIZE AN IMENSE VARIETY OF POSSIBLE PATTERNS RELATED TO THEYR SENSORIAL ORGANS. If we are talking about the human brain, we can see that virtualy any visual pattern can be recognized, any sound pattern can be recognized, any movement pattern, activity pattern, verbal pattern, social pattern, etc.
The standard neural network model gives a good explanation of how the pattern recognition functions. Even if the current technology knowledge does not allow us to build artificial systems of pattern recognition as powerfull as the biological one, I consider it is only a technological limitation, not a theoretical one, as the same that we consider that inter-stelar travel is only a technological limitation, not a fundamental one. On the other side, traveling with velocities grater then the velocity of light is a theorical impossibity according to current physics.
Any representation framework is composed of a set of caracteristics (space, time, energy, frequence, color, etc), so it represents a set of possible patterns available to the representation framework (or you can use the term representation space). What I’m proposing is that if the brain is capable of recognize patterns, then it is very resonamble to suggest that the brain can recognize any pattern of any representation space. Of course, many representation spaces is not obivously recognized, needing cultural learning and development to be recognized due to its degree of complexity, as are the more advanced cultural concepts.
So the answer of your question is that I argue that in theory for any representation space needed to describe a given pattern, we can project a pattern recognition system capable to recognize that pattern. This is what seems to be the case of the human brain. At least all that human brain can recognize belongs to the representation space available to our pattern recognition brain-mind-sensorial system.
I wish I could be clear to you without a specific exemple,
Gilberto de Paiva -
Dear chris lofting,
I tried to read your work on IDM, but its a little extensive reading for a first talk. But what I can understand from your post is that you are trying to make a detailed mind functioning modeling according to your IDM model. I have to say that I’m not able to agree or disagree at first glance. I have to think more about, then I can give you my avaliation of your work.
But your modeling approach is to explain the detailed mechanisms, which difer from my modeling approach is to explain some general mechanisms.
My modeling strategy is to describe the mind functioning with a general theory as is Darwing Evolution Theory. Darwing did not tryied to explain the details of the mechanims of the genetics and DNA/RNA (he not even knew it at his time). Not even Darwing tryied to estabilish the detailed and acurated natural selection paths that exactly occured.
For my point of view, Darwing proposed the general principles that could explain or describe by natural simple general mechanisms the evolution of the living species. That is what exactly I’m proposing, with the PATTERN RECOGNITION principle in my theory been analogous the the NATURAL SELECTION principle in Darwings theory.
I know it is a little ambicious for my part …
But I believe I’m in the correct way …
The future will show us if more detailed or more general theoretical strategies (or a mix of them) can give us a better theory of mind.
Thank You for you attention and interest,
Gilberto de Paiva -
The IDM work is GENERAL in that it covers the foundations of language and so all that we can communicate from a VAGUE position where local context then customises through use of symbols. Darwinian principles apply to local context development and cover the stability of the neuron where the entymology goes back some 600 million years.
Given LOCAL context Darwinian principles, the adaptation to context has included the utilisation of neurons for information processing and as such the filtering system of all meaning. Since the METHOD used to derive meaning determines what is meaningful, so all possible meanings can be identified by consideration of that method.
In our brains the neurology demonstrates information gathering etc through recursion of the differentiate/integrate dichotomy where such elicits a dimension of patterns, classes of meanings, available for all neuron-dependent life forms. (see refs etc in the IDM page or in such pdf documents as Categories of Mediation )
DEPTH in recursion is the source of language and so the ability for neuron-dependent life forms like us to describe ourselves by reference to ourselves through the use of analogy/metaphor – aka pattern-matching ;-)
Chris.
-
Ok chris lofting,
I really need some time to think about your ideas to give you a better answer. Give me few days, I will read your work more carefully and I will answer you.
Gilberto
-
Dear Gilberto: Please take a look at Steve Grossberg´s paper published in Consciousness & Cognition (check the Forum “Stephen Grossberg Consciousness Researcher” here in BPCC). If you cannot get the PDF please e-mail me privately.
Best
Alfredo -
Dear Gilberto,
if you look at recognition from an evolutionary perspective, I would point to most probably earlier mechanisms that are not related to “mind” but to perception.
Yours friendly
Hans -
Hi Alfredo, thanks a lot for introducing the ART theory from Grossberg to me.
I read some recent papers from him and one that exat explains the ART theory:Grossberg, S. (2003). Adaptive resonance theory. Technical Report CAS/CNS TR-2000-024 , Boston University . In The encyclopedia of cognitive science . London : Macmillan Reference Ltd. Available in PDF (http://cns.bu.edu/~steve/Gro2003EncyclopediaCogSci.pdf)(3,567Kb).
I separated some parts of this paper to discuss with you, and also I will email Grossberg to try to discuss with him (do you have contact to him? His line of thinking interested me.):
Grossberg modeling theory is focused at theoretical artificial neural networks possible mechanisms. I didn’t see any pratical or software implementation or testing of an ART based neural network, as I could see from the Grosberg papers I read, if you know any I would really like to read about. Grossberg also discuss the relations and implications of ART theory to biological nerological systems, but as I could see, it is not his main focus.
ART theory proposes a set of neural networks mechanisms at some detailed level (not a pratical software implementation level as I could understand) to model some proposed neural funcionalities like feedback, ressonance reinforcement, memorization, basic learning, etc, but I see some limitations at the proposals of higher cognition functions and concepts as consciousness definition.
My theoretical modeling approach differs from his at this point, that the pattern recognition and processing mechanism also produces definitions and descriptions of some basic cognitive functions, but the main quality of my theory is that it gives the definition of higher cognitive functions in a more comprehensive, simple and structured than Grossberg does.
Let’s compare some parts if we ask as how a physical system can be conscious of its own mental activities:
page 7 of Grossberg paper proposal for a definition of consciousness:
“The resonance between these two types of information converts the pattern of attended features into a coherent context-sensitives tatet hat is linked to its category through feedback. It is this coherents tate, that joins togetherd istributed featuresa nd symbolic categories,t hat can enter consciousnessA. RT
predicts that all conscious states are resonant states. In particular, such a resonance binds spatially distributed features into either a stable equilibrium or a synchronous oscillation.”My proposal for a definition of conscioussnes:
“A general pattern recognition mechanism is a system that can recognize any pattern of sensation, visual stimulus, audio pattern, activity pattern, concept pattern, etc. If this system can memorize and process these recognized patterns in a thinking process they are conscious for my point of view.”One derivation my definition is the easy definition of what means to been conscious of itself mind and consciousness, which is simple the recognition of the patterns of its own mental activity. I can’t see how Grosberg definition can deal with this aspect of conciousness of one’s own mind and consciousness.
Some parts of Grossberg paper have equivalent argumentation as mine:
page 9 of Grossberg paper
“It has also been shown that the adaptive weights which are learned by some ART models can, at any stage ofleaming, be translated into IF-THEN rules (e.g., Carpenter and Grossberg, 1994). Thus the ART model is a self-organizing rule-discovering production system as well as a neural network. These examples show that the claims of some cognitive scientists and AI practioners that neural network models cannot learn rule-based behaviors are incorrect.”I proposed that the patterns recognized and processed by the mind, I called mental patterns, must be somehow activated for the thinking processing, and one simple mechanism of acitivation I proposed is the association of mental patterns, which is exactly the logical statement that If mental pattern X is activated then mental pattern Y will be activated for association. For me, Grossberg conclusion and my proposal are equivalent. The diference is that Grossberg arrived to his conclusion from the properties of the ART neural networks, and for me it is only a proposed functionality.
So Alfredo,
if you have more profound knowledge on ART theory, I’d like to discuss with you in advance.Gilberto
Results
-
