Abstract
The brain is without doubt the most complex adaptive system known to humanity, arguably also a complex system about which we know very little.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
See, e.g., Abeles et al. (1995) and Kenet et al. (2003).
- 2.
Humans can distinguish cognitively about 10–12 objects per second.
- 3.
See Edelman and Tononi (2000).
- 4.
- 5.
See Crick and Koch (2003).
- 6.
Note that neuromodulators are typically released in the intercellular medium from where they physically diffuse towards the surrounding neurons.
- 7.
This is a standard result for so-called Hopfield neural networks, see e.g. Ballard (2000).
- 8.
A neural network is denoted “recurrent” when loops dominate the network topology.
- 9.
For a mathematically precise definition, a memory is termed fading when forgetting is scale-invariant, viz having a power law functional time dependence.
- 10.
From the point of view of dynamical systems theory effective freedom of action is conceivable in connection to a true dynamical phase transition, like the ones discussed in the Chap. 4 possibly occurring in a high-level cognitive system. Whether dynamical phase transitions are of relevance for the brain of mammals, e.g. in relation to the phenomenon of consciousness, is a central and yet unresolved issue.
- 11.
We note that general n-point interactions could be generated additionally when eliminating the interneurons. “n-point interactions” are terms entering the time evolution of dynamical systems depending on \((n-1)\) variables. Normal synaptic interactions are 2-point interactions, as they involve two neurons, the presynaptic and the postsynaptic neuron. When integrating out a degree of freedom, like the activity of the interneurons, n-point interactions are generated generally. The postsynaptic neuron is then influenced only when \((n-1)\) presynaptic neurons are active simultaneously. n-point interactions are normally not considered in neural networks theory. They complicate the analysis of the network dynamics considerably.
- 12.
Here we use the term “transient attractor” as synonymous with “attractor ruin”, an alternative terminology from dynamical system theory.
- 13.
A possible mathematical implementation for the reservoir functions, with \(\alpha={\textrm{w}},z\), is \( f_\alpha(\varphi)\ =\ f_\alpha^{(\min)} \,+\, \left(1-f_\alpha^{(\min)}\right) \frac{ \textrm{atan}[(\varphi-\varphi_c^{(\alpha)})/\varGamma_\varphi] - \textrm{atan}[(0-\varphi_c^{(\alpha)})/\varGamma_\varphi]}{\textrm{atan}[(1-\varphi_c^{(\alpha)})/\varGamma_\varphi] - \textrm{atan}[(0-\varphi_c^{(\alpha)})/\varGamma_\varphi] } \). Suitable values are \(\varphi_c^{(z)}=0.15\), \(\varphi_c^{({\textrm{w}})}=0.7\) \(\varGamma_\varphi=0.05\), \(f_{\textrm{w}}^{(\min)}=0.1\) and \(f_z^{(\min)}=0\).
- 14.
A Kohonen network is an example of a neural classifier via one-winner-takes-all architecture, see e.g. Ballard (2000).
Further Reading
For a general introduction to the field of artificial intelligence (AI), see Russell (1995). For a handbook on experimental and theoretical neuroscience, see Arbib (2002). For exemplary textbooks on neuroscience, see Dayan (2001) and for an introduction to neural networks, see Ballard (2000).
Somewhat more specialized books for further reading regarding the modeling of cognitive processes by small neural networks is that by McLeod et al. (1998) and on computational neuroscience that by O’Reilly (2000).
For some relevant review articles on dynamical modeling in neuroscience the following are recommended: Rabinovich et al. (2006); on reinforcement learning Kaelbling et al. (1996), and on learning and memory storage in neural nets Carpenter (2001).
We also recommend to the interested reader to go back to some selected original literature dealing with “simple recurrent networks” in the context of grammar acquisition (Elman, 990, 2004), with neural networks for time series prediction tasks (Dorffner, 1996), with “learning by error” (Chialvo and Bak, 1999), with the assignment of the cognitive tasks discussed in Sect. 8.3.1 to specific mammal brain areas (Doya, 1999), with the effect on memory storage capacity of various Hebbian-type learning rules (Chechik et al., 2001), with the concept of “associative thought processes” (Gros, 2007, 2009a) and with “diffusive emotional control” (Gros, 2009b).
It is very illuminating to take a look at the freely available databases storing human associative knowledge (Nelson et al., 1998; Liu, 2004).
Abeles M. et al. 1995 Cortical activity flips among quasi-stationary states. Proceedings of the National Academy of Science, USA 92, 8616–8620.
Arbib, M.A. 2002 The Handbook of Brain Theory and Neural Networks. MIT Press, Cambridge, MA.
Baars, B.J., Franklin, S. 2003 How conscious experience and working memory interact. Trends in Cognitive Science 7, 166–172.
Ballard, D.H. 2000 An Introduction to Natural Computation. MIT Press, Cambridge, MA.
Carpenter, G.A. 2001 Neural-network models of learning and memory: Leading questions and an emerging framework. Trends in Cognitive Science 5, 114–118.
Chechik, G., Meilijson, I., Ruppin, E. 2001 Effective neuronal learning with ineffective Hebbian learning rules. Neural Computation 13, 817.
Chialvo, D.R., Bak, P. 1999 Learning from mistakes. Neuroscience 90, 1137–1148.
Crick, F.C., Koch, C. 2003 A framework for consciousness. Nature Neuroscience 6, 119–126.
Dayan, P., Abbott, L.F. 2001 Theoretical Neuroscience: Computational and Mathematical Modeling of Neural Systems. MIT Press, Cambridge, MA.
Dehaene, S., Naccache, L. 2003 Towards a cognitive neuroscience of consciousness: Basic evidence and a workspace framework. Cognition 79, 1–37.
Dorffner, G. 1996 Neural networks for time series processing. Neural Network World 6, 447–468.
Doya, K. 1999 What are the computations of the cerebellum, the basal ganglia and the cerebral cortex? Neural Networks 12, 961–974.
Edelman, G.M., Tononi, G.A. 2000 A Universe of Consciousness. Basic Books, New York.
Elman, J.L. 1990 Finding structure in time. Cognitive Science 14, 179–211.
Elman, J.L. 2004 An alternative view of the mental lexicon. Trends in Cognitive Sciences 8, 301–306.
Gros, C. 2007 Neural networks with transient state dynamics. New Journal of Physics 9, 109.
Gros, C. 2009a Cognitive computation with autonomously active neural networks: An emerging field. Cognitive Computation 1, 77.
Gros, C. 2009b Emotions, diffusive emotional control and the motivational problem for autonomouscognitive systems. In: Vallverdu, J., Casacuberta, D. (eds.), Handbook of Research on Synthetic Emotionsand Sociable Robotics: New Applications in Affective Computing and Artificial Intelligence. IGI-Global Hershey, NJ
Kaelbling, L.P., Littman, M.L., Moore, A. 1996 Reinforcement learning: A survey. Journal of Artificial Intelligence Research 4, 237–285.
Kenet, T., Bibitchkov, D., Tsodyks, M., Grinvald, A., Arieli, A. 2003 Spontaneously emerging cortical representations of visual attributes. Nature 425, 954–956.
Liu, H., Singh, P. 2004 ConcepNet a practical commonsense reasoning tool-kit. BT Technology Journal 22, 211–226.
McLeod, P., Plunkett, K., Rolls, E.T. 1998 Introduction to Connectionist Modelling. Oxford University Press, New York.
Nelson, D.L., McEvoy, C.L., Schreiber, T.A. 1998 The University of South Florida Word Association, Rhyme, and Word Fragment Norms. Homepage: http://www.usf.edu/FreeAssociation.
O’Reilly, R.C., Munakata, Y. 2000 Computational Explorations in Cognitive Neuroscience: Understanding the Mind by Simulating the Brain. MIT Press, Cambridge.
Rabinovich, M.I., Varona, P., Selverston, A.I., Abarbanel, H.D.I. 2006 Dynamical principles in neuroscience. Review of Modern Physics 78, 1213–1256.
Russell, S.J., Norvig, P. 1995 Artificial Intelligence: A Modern Approach. Prentice-Hall, Englewood Cliffs, NJ.
Author information
Authors and Affiliations
Corresponding author
Exercises
Exercises
8.1.1 Transient State Dynamics
Consider a system containing two variables, \(x,\varphi\in[0,1]\). Invent a system of coupled differential equations for which \(x(t)\) has two transient states, \(x\approx1\) and \(x\approx0\). One possibility is to consider ϕ as a reservoir and to let \(x(t)\) autoexcite/autodeplete itself when the reservoir is high/low.
The transient state dynamics should be rigorous. Write a code implementing the differential equations.
8.1.2 The Diffusive Control Unit
Given are two signals \(y_1(t)\in[0,\infty]\) and \(y_2(t)\in[0,\infty]\). Invent a system of differential equations for variables \(x_1(t)\in[0,1]\) and \(x_2(t)\in[0,1]\) driven by the \(y_{1,2}(t)\) such that \(x_1\to1\) and \(x_2\to0\) when \(y_1>y_2\) and vice versa. Note that the \(y_{1,2}\) are not necessarily normalized.
8.1.3 Leaky Integrator Neurons
Consider a two-site network of neurons, having membrane potentials x i and activities \(y_i\in[-1,1]\), the so-called “leaky integrator” model for neurons,
with \(\varGamma>0\) being the decay rate. The coupling \({\textrm{w}}>0\) links neuron one (two) excitatorily (inhibitorily) to neuron two (one). Which are the fixpoints and for which parameters can one observe weakly damped oscillations?
8.1.4 Associative Overlaps and Thought Processes
Consider the seven-site network of Fig. 8.6. Evaluate all pairwise associative overlaps of order zero and of order one between the five cliques, using Eqs. (8.4) and (8.5). Generate an associative thought process of cliques \(\alpha_1,\ \alpha_2,\ldots\), where a new clique \(\alpha_{t+1}\) is selected using the following simplified dynamics:
-
(1)
\({\alpha_{t+1}}\) has an associative overlap of order zero with \({\alpha_{t}}\) and is distinct from \({\alpha_{t-1}}\).
-
(2)
If more than one clique satisfies criterium (1), then the clique with the highest associative overlap of order zero with \(\alpha_{t}\) is selected.
-
(3)
If more than one clique satisfies criteria (1)–(2), then one of them is drawn randomly.
Discuss the relation to the dHAN model treated in Sect. 8.4.2.
Rights and permissions
Copyright information
© 2011 Springer-Verlag Berlin Heidelberg
About this chapter
Cite this chapter
Gros, C. (2011). Elements of Cognitive Systems Theory. In: Complex and Adaptive Dynamical Systems. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-04706-0_8
Download citation
DOI: https://doi.org/10.1007/978-3-642-04706-0_8
Published:
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-642-04705-3
Online ISBN: 978-3-642-04706-0
eBook Packages: Physics and AstronomyPhysics and Astronomy (R0)