ArticlePDF Available

The Correlation Theory of Brain Function

Authors:
  • Frankfurt Institute for Advanced Studies

Abstract

A summary of brain theory is given so far as it is contained within the framework of Localization Theory. Diffculties of this "conventional theory" are traced back to a specific deficiency: there is no way to express relations between active cells (as for instance their representing parts of the same object). A new theory is proposed to cure this deficiency. It introduces a new kind of dynamical control, termed synaptic modulation, according to which synapses switch between a conducting and a non- conducting state. The dynamics of this variable is controlled on a fast time scale by correlations in the temporal fine structure of cellular signals. Furthermore, conventional synaptic plasticity is replaced by a refined version. Synaptic modulation and plasticity form the basis for short-term and long-term memory, respectively. Signal correlations, shaped by the variable network, express structure and relationships within objects. In particular, the figure-ground problem may be solved in this way. Synaptic modulation introduces flexibility into cerebral networks which is necessary to solve the invariance problem. Since momentarily useless connections are deactivated, interference between different memory traces can be reduced, and memory capacity increased, in comparison with conventional associative memory.
S
˜1S
˜1
a)b)c)
S
˜1S
˜2
+
615
9
12
8
4
2
10
3
7
11
4
8
12
9
5
1
6
11
7
3
10 2
6
5
4
3
2
1
12
11
10
9
7
8
... This is in line with the previous findings that neural networks fall short in generalization. One hypothesis for this is due to their inability to dynamically and flexibly bind information distributed throughout the network [17], also known as the binding problem [18], [19]. To address this issue, there has been an increased interest in designing object-centric neural networks that learn (discrete) object representations from raw visual input, which support efficient learning and generalization to novel scenarios and behaviors. ...
... One potential explanation of the poor generalization of neural networks is that they cannot dynamically and flexibly bind the information distributed throughout the network [17]. This is also known as the binding problem [18], [19]. Recent work addresses this through object-centric neural networks that learn (discrete) object representations from raw visual input, to support efficient learning and generalization to novel scenarios and behaviors. ...
Preprint
Full-text available
Reinforcement learning agents must generalize beyond their training experience. Prior work has focused mostly on identical training and evaluation environments. Starting from the recently introduced Crafter benchmark, a 2D open world survival game, we introduce a new set of environments suitable for evaluating some agent's ability to generalize on previously unseen (numbers of) objects and to adapt quickly (meta-learning). In Crafter, the agents are evaluated by the number of unlocked achievements (such as collecting resources) when trained for 1M steps. We show that current agents struggle to generalize, and introduce novel object-centric agents that improve over strong baselines. We also provide critical insights of general interest for future work on Crafter through several experiments. We show that careful hyper-parameter tuning improves the PPO baseline agent by a large margin and that even feedforward agents can unlock almost all achievements by relying on the inventory display. We achieve new state-of-the-art performance on the original Crafter environment. Additionally, when trained beyond 1M steps, our tuned agents can unlock almost all achievements. We show that the recurrent PPO agents improve over feedforward ones, even with the inventory information removed. We introduce CrafterOOD, a set of 15 new environments that evaluate OOD generalization. On CrafterOOD, we show that the current agents fail to generalize, whereas our novel object-centric agents achieve state-of-the-art OOD generalization while also being interpretable. Our code is public.
... Neurones with the same firing phase belong to the same auditory objects. In 1976 and 1981, the temporal correlation that performs binding was proposed by Milner [54] and independently by Malsburg [55,56,57]. Milner and Malsburg have observed that synchrony is a crucial feature to bind neurones that are associated with similar characteristics. ...
... It combines a reconstruction analysis/synthesis cochlear filterbank along with auditory image representations of audible signals with a spiking neural network. The segregation and binding of the auditory objects (coming from different sound sources) is performed by the spiking neural network (implementing the temporal correlation [54,55]) that also generates a mask 7 to be used in conjunction with the synthesis filterbank to generate the separated sound sources. The neural network uses third generation neural networks, where neurones are usually called spiking neurones [43]. ...
Preprint
Source separation and speech recognition are very difficult in the context of noisy and corrupted speech. Most conventional techniques need huge databases to estimate speech (or noise) density probabilities to perform separation or recognition. We discuss the potential of perceptive speech analysis and processing in combination with biologically plausible neural network processors. We illustrate the potential of such non-linear processing of speech on a source separation system inspired by an Auditory Scene Analysis paradigm. We also discuss a potential application in speech recognition.
... The space dimension is used in the representation of incoming sensory information by means of topographically organized (for instance retinotopic or tonotopic) neural activity, thus giving rise to the distributed nature of cortical activity (Quiroga et al., 2005). The time dimension is used as a way to establish brief reciprocal connections between distant populations (von der Malsburg, 1994;Varela, 1995;Varela et al., 2001) in periods as brief as 10 ms, as suggested by behavioral (Goodale et al., 1986) and electrophysiology experiments (Maldonado et al., 2008). These are the two dimensions that the brain uses to bind and segregate features and to move from the current to the next brain state. ...
Article
Full-text available
Bi-stable perception is a strong instance of cognitive self-organization, providing a research model for how ‘the brain makes up its mind.’ The complexity of perceptual bistability prevents a simple attribution of functions to areas, because many cognitive processes, recruiting multiple brain regions, are simultaneously involved. The functional magnetic resonance imaging (fMRI) evidence suggests the activation of a large network of distant brain areas. Concurrently, electroencephalographic and magnetoencephalographic (MEEG) literature shows sub second oscillatory activity and phase synchrony on several frequency bands. Strongly represented are beta and gamma bands, often associated with neural/cognitive integration processes. The spatial extension and short duration of brain activities suggests the need for a fast, large-scale neural coordination mechanism. To address the range of temporo-spatial scales involved, we systematize the current knowledge from mathematical models, cognitive sciences and neuroscience at large, from single-cell- to system-level research, including evidence from human and non-human primates. Surprisingly, despite evidence spanning through different organization levels, models, and experimental approaches, the scarcity of integrative studies is evident. In a final section of the review we dwell on the reasons behind such scarcity and on the need of integration in order to achieve a real understanding of the complexities underlying bi-stable perception processes.
Article
Full-text available
Did the boy bite the cat or was it the other way around? When processing a sentence with several objects, one has to establish ‘who did what to whom’. When a sentence cannot be interpreted by recalling an image from memory, we rely on the special type of voluntary constructive imagination called Prefrontal synthesis (PFS). PFS is defined as the ability to juxtapose mental visuospatial objects at will. We hypothesised that PFS has fundamental importance for language acquisition. To test this hypothesis, we designed a PFS-targeting intervention and administered it to 6,454 children with language deficiencies (age 2 to 12 years). The results from the three-year-long study demonstrated that children who engaged with the PFS intervention showed 2.2-fold improvement in combinatorial language comprehension compared to children with similar initial evaluations. These findings suggest that language can be improved by training the PFS and exposes the importance of the visuospatial component of language. This manuscript reflects on the experimental findings from the point of view of human language evolution. When used as a proxy for evolutionary language acquisition, the study results suggest a dichotomy of language evolution, with its speech component and its visuospatial component developing in parallel. The study highlights the radical idea that evolutionary acquisition of language was driven primarily by improvements of voluntary imagination rather than by improvements in the speech apparatus.
Book
Full-text available
Human aesthetic experience: How does aesthetic evaluation work in terms of psychological and neurobiological evidence? How do we assign meaning to sensual percepts? Why do we find commonalities as well as differences between individual aesthetic perception, interpretation and evaluation? New models and categorization structures are suggested.
Article
Though much progress has been made to understand feature integration, debate remains regarding how objects are represented in mind based on their constituent features. Here, we advance this debate by introducing a novel shape-color “conjunction task” to reconstruct memory resolution for multiple object features simultaneously. In a first experiment, we replicate and extend a classic paradigm originally tested using a change detection task. Replicating previous work, memory resolution for individual features was reduced when the number of objects increased, regardless of the number of to-be-remembered features. Extending previous work, we found that high resolution memory near perfect in resemblance to the target was selectively impacted by the number of to-be-remembered features. Applying a data-driven statistical model of stochastic dependence, we found robust evidence of integration for lower-resolution feature memories, but less evidence for integration of high-resolution feature memories. These results suggest that memory resolution for individual features can be higher than memory resolution for their integration. In a second experiment which manipulated the nature of distracting information, we examined whether object features were directly bound to each other or by virtue of shared spatial location. Feature integration was disrupted by distractors sharing visual features of target objects but not when distractors shared spatial location – suggesting that feature integration can be driven by direct binding between shape and color features in memory. Our results constrain theoretical models of object representation, providing empirical support for hierarchical representations of both integrated and independent features.
Article
Full-text available
A newly discovered physical mechanism involving incoherent electron tunneling in layers of the protein ferritin that are found in catecholaminergic neurons (catecholaminergic neuron electron transport or CNET) is hypothesized to support communication between neurons. Recent tests further confirm that these ferritin layers can also perform a switching function (in addition to providing an electron tunneling mechanism) that could be associated with action selection in those neurons, consistent with earlier predictions based on CNET. While further testing would be needed to confirm the hypothesis that CNET allows groups of neurons to communicate and act as a switch for selecting one of the neurons in the group to assist in reaching action potential, this paper explains how that hypothesized behavior would be consistent with Integrated Information Theory (IIT), one of a number of consciousness theories (CTs). While the sheer number of CTs suggest that any one of them alone is not sufficient to explain consciousness, this paper demonstrates that CNET can provide a physical substrate and action selection mechanism that is consistent with IIT and which can also be applied to other CTs, such as to conform them into a single explanation of consciousness.
Article
Full-text available
Gamma-band activity, peaking around 30–100 Hz in the local field potential's power spectrum, has been found and intensively studied in many brain regions. Although gamma is thought to play a critical role in processing neural information in the brain, its cognitive functions and neural mechanisms remain unclear or debatable. Experimental studies showed that gamma rhythms are stochastic in time and vary with visual stimuli. Recent studies further showed that multiple rhythms coexist in V1 with distinct origins in different species. While all these experimental facts are a challenge for understanding the functions of gamma in the visual cortex, there are many signs of progress in computational studies. This review summarizes and discusses studies on gamma in the visual cortex from multiple perspectives and concludes that gamma rhythms are still a mystery. Combining experimental and computational studies seems the best way forward in the future.
Book
Full-text available
In psychotherapy and psychiatry, mental illness diagnostics is limited to the description of symptoms and has nothing to do with assessing the state of the organ that performs the mental functions. In other branches of medicine, any diagnosis aims to answer the question about the physical disturbance of the internal technology of the body. But the Mind remains outside of this correct approach. As a result, all systemic mental pathologies are currently incurable. This absurd situation is due to the lack of a physically grounded theoretical model of the Mind. Without it, practical treatment of the disorders remains ‘wandering in the dark.’ How can we fix something if there is no understanding of what it is and how it works? Within the Teleological Transduction Theory that the author consistently developed in the previous volumes of the series, harmony of the Mind is a physical concept of normal functioning. Respectively, when he describes the pathologies as dissonances, it is not a metaphor but a reference to the physical and technological malfunction leading to the disintegration of consciousness. The author does not aim to describe all pathologies in this book. His mission is to create a concept as a universal tool that can be used in different situations and various ways. The author suggests an entirely novel approach to analysing, diagnosing, and classifying mental pathologies. As an example, he develops a model of such systemic pathologies as autism and schizophrenia, uncovering the essence of the underlying problem that causes the observed symptoms. This physical, physiological and technological explanation of the states of chronic dissonance of the Mind may be the beginning of the transition to the actual diagnosis and treatment of mental illnesses as brain disorders.
Preprint
A newly-discovered physical mechanism involving electron tunneling in layers of the protein ferritin that are found in catecholaminergic neurons (catecholaminergic neuron electron transport or CNET), is hypothesized to support communication between neurons. Recent tests further confirm that these ferritin layers can also perform a switching function (in addition to providing an electron tunneling mechanism) that could be associated with action selection in those neurons, consistent with earlier predictions based on CNET. While further testing would be needed to confirm the hypothesis that CNET allows groups of neurons to communicate and act as a switch for selecting one of the neurons in the group to assist in reaching action potential, this paper explains how that hypothesized behavior would be consistent with Integrated Information Theory (IIT), one of a number of consciousness theories (CTs). While the sheer number of CTs suggest that any one of them is not sufficient to explain consciousness, this paper demonstrates that CNET can provide a physical substrate that is consistent with IIT and which can also be applied to other CTs, such as to conform them into a single explanation of consciousness.
Article
Full-text available
We present a method for setting up topographically ordered mappings between two sheets of nerve cells. A set of chemical markers that express the neighborhood relationships within the presynaptic sheet is induced by the fibers into the postsynaptic sheet. The markers are used to guide the fibers to their terminal sites. A case for which this idea may be relevant is the retinotectal projection; our model exhibits types of plasticity found experimentally. The fact that the postsynaptic markers remain after removal of the projecting fibers suggests an important difference between development and regeneration. This paper concentrates on explaining the basic idea, and in addition presents a set of preliminary computer simulations.
Article
Full-text available
The problem discussed is the relationship between the firing of single neurons in sensory pathways and subjectively experienced sensations. The conclusions are formulated as the following five dogmas: To understand nervous function one needs to look at interactions at a cellular level, rather than either a more macroscopic or microscopic level, because behaviour depends upon the organized pattern of these intercellular interactions. The sensory system is organized to achieve as complete a representation of the sensory stimulus as possible with the minimum number of active neurons. Trigger features of sensory neurons are matched to redundant patterns of stimulation by experience as well as by developmental processes. Perception corresponds to the activity of a small selection from the very numerous high-level neurons, each of which corresponds to a pattern of external events of the order of complexity of the events symbolized by a word. High impulse frequency in such neurons corresponds to high certainty that the trigger feature is present. The development of the concepts leading up to these speculative dogmas, their experimental basis, and some of their limitations are discussed.
Article
Cells having some properties similar to those of neurons are considered. A mass of such cells, randomly placed together with a uniform volume density, appears capable of supporting various simple forms of activity, including plane waves, spherical and circular waves and vortex effects. The propagation of a plane wave of activity has been considered in some detail. It is shown that a wave may be initiated in a mass of such cells by a number of individual stimuli. The mass has a very sensitive threshold to such stimulation. This threshold depends on cell properties, and by altering the threshold a mass of cells may be made to act as an on/off switch. The switching of waves may be compared with the shifting of attention in a living organism. Particularly interesting phenomena emerge if some property of the individual cells, e.g. size, extent of axon or dendrite structure, or threshold, changes with repeated use. The mass of cells may then exhibit an ability to modify its response according to past experience in a manner similar to that of living organisms. Trial and error learning, conditioned responses, and the ability to regenerate internally a sequence of past events may be demonstrated with very little complication of the form of the mass of cells.
Article
The assumption is explored that the effect of information on the brain is to cause the neurons to form groups. Each group can “ignite” to reverberatory firing in response to a certain kind of stimulation, similar to that which originally formed it. The groups have to be larger than a certain critical size in order to be able to ignite. The critical size is expressible in terms of the number of branches per neuron and the number of neurons in the brain; it is found to be around 105–106 neurons. It is shown that about 107–109 groups can be fitted into the brain before they begin to interact significantly. (One neuron can participate in many groups.) We assume that the latter number of groups is something like the maximum number of elementary “things” that the brain can “know”. Attention is devoted to such questions as reliability, interaction, and the mechanism of ignition, and to possible indications that the brain is actually organized in the scheme described.
Article
We show that given certain plausible assumptions the existence of persistent states in a neural network can occur only if a certain transfer matrix has degenerate maximum eigenvalues. The existence of such states of persistent order is directly analogous to the existence of long range order in an Ising spin system; while the transition to the state of persistent order is analogous to the transition to the ordered phase of the spin system. It is shown that the persistent state is also characterized by correlations between neurons throughout the brain. It is suggested that these persistent states are associated with short term memory while the eigenvectors of the transfer matrix are a representation of long term memory. A numerical example is given that illustrates certain of these features.
An important problem in biology is to explain how patterned neural connections are set up during ontogenesis. Topographically ordered mappings, found widely in nervous systems, are those in which neighbouring elements in one sheet of cells project to neighbouring elements in a second sheet. Exploiting this neighbourhood property leads to a new theory for the establishment of topographical mappings, in which the distance between two cells is expressed in terms of their similarity with respect to certain physical properties assigned to them. This topographical code can be realized in a model employing either synchronization of nervous activity or exchange of specific molecules between neighbouring cells. By means of modifiable synapses the code is used to set up a topographical mapping between two sheets with the same internal structure. We have investigated the neural activity version. Without needing to make any elaborate assumptions about its structure or about the operations its elements are to carry out we have shown that the mappings are set up in a system-to-system rather than a cell-to-cell fashion. The pattern of connections develops in a step-by-step and orderly fashion, the orientation of the mappings being laid down in the earliest stages of development.
Article
This paper examines the idea that ordered patterns of nerve connections are set up by means of markers carried by the individual cells. The case of the ordered retinotectal projection in amphibia and fishes is discussed in great detail. It is suggested that retinotectal mappings are the result of two mechanisms acting in concert. One mechanism induces a set of retinal markers into the tectum. By this means, an initially haphazard pattern of synapses is transformed into a continuous or piece-wise continuous projection. The other mechanism places the individual pieces of the map in the correct orientation. The machinery necessary for this inductive scheme has been expressed in terms of a set of differential equations, which have been solved numerically for a number of cases. Straightforward assumptions are made as to how markers are distributed in the retina; how they are induced into the tectum; and how the induced markers bring about alterations in the pattern of synaptic contacts. A detailed physiological interpretation of the model is given. The inductive mechanism has been formulated at the level of the individual synaptic interactions. Therefore, it is possible to specify, in a given situation, not only the nature of the end state of the mapping but also how the mapping develops over time. The role of the modes of growth of retina and tectum in shaping the developing projection becomes clear. Since, on this model, the tectum is initially devoid of markers, there is an important difference between the development and the regeneration of ordered mappings. In the development of duplicate maps from various types of compound-eyes, it is suggested that the tectum, rather than the retina, contains an abnormal distribution of markers. An important parameter in these experiments, and also in the regeneration experiments where part-duplication has been found, is the range of interaction amongst the retinal cells. It is suggested that the results of many of the regeneration experiments (including apparently contradictory ones) are manifestations of a conflict between the two alternative ways of specifying the orientation of the map: through the information carried by the markers previously induced into the tectum and through the orientation mechanism itself.
This paper lists twelve conditions that must be fulfilled by a satisfactory theory of visual pattern recognition in animals and man: such a theory must explain (1) size invariance, (2) position invariance, (3) brightness invariance, (4) the equivalence of outline and filled-in shapes, (5) lack of invariance under most rotations, (6) the known confusions made between patterns, (7) animals' ability to disregard jitter, (8) the human ability to segment shapes in different ways, (9) the way in which man recognizes complex scenes without recognizing individual details of the scene, (10) perceptual learning, (11) the way in which the brain takes advantage of the redundancy of the visual environment; (12) finally any theory must be consistent with the physiological evidence. The outlines of a theory meeting these conditions are put forward. The model has three parts: (1) A processor that extracts local features from the input picture preserving information about the spatial relationships between the features. (2) A mechanism that induces an abstract description of the output from the processor. (3) A store in which such descriptions are held. Some suggestions are made about the language in which descriptions are framed: it must contain hierarchical elements to allow for the possibility of segmenting a picture in different ways. It is shown how the properties of such a language could account for the facts listed above. In recognizing a picture the output from the store is matched to a stored descriptive rule. What we see is the descriptive rule selected to describe an input picture-we cannot respond to details of the picture not represented in the rule. Since, however, many different rules describe the same input picture, animals and man respond to different aspects of the same picture on different occasions of presentation.