The connection machine
The BSB model: a simple nonlinear autoassociative neural network
Associative neural memories
Mapping hierarchical neural networks to VLSI hardware
Neural Networks
Weight-space mapping of fMRI language tasks
CNS '97 Proceedings of the sixth annual conference on Computational neuroscience : trends in research, 1998: trends in research, 1998
Reconfigurable networking for coordinated multi-agent sensing and communications
Information Sciences—Applications: An International Journal
Hi-index | 0.00 |
The neocortex is arguably the most sophisticated structure within the mammalian brain. It is the largest brain structure in the human, and its properties endow us with qualities that are unique to our species. In order to develop a systems-level approach to understanding the neocortex, and to develop a theoretical basis that is both tractable and useful, enormous anatomical and physiological simplifications must be made. These include focusing on specific aspects of function, such as associative memory and learning, at the expense of many other important characteristics. Such simplifications are required in any systems-level model of the brain.The main contribution of this chapter is to describe a model of the neocortex that links together different scales or levels of neural organization. Specifically, individual neurons cluster together into networks, and these networks cluster together into larger networks, and so on. We argue that dynamically reconfigurable networks of neurons exist at multiple scales, and are fundamental to the structural and functional integrity of the cortex.Experimental data supporting both the approach and the predictions of our neocortical model, termed the Network of Networks (NoN), are described. Section 7.2 summarizes some of the neurobiology relevant to the theoretical approach. The NoN is described in Section 7.3, where it is suggested that the model should do more than simply provide descriptors of neocortical organization function. Instead, it should lead to new ways of understanding and applying rapid, parallel and associative computations within and between neural networks at different scales. This is discussed in Section 7.4, where we investigate the model's veracity in tests of predicatability and falsifiability. In the final section, implications of the NoN for neuroengineering are considered.