Oscillatory Neural Networks: Modeling Binding and Attention by Synchronization of Neural Activity
This chapter presents some mathematical models in support of the hypothesis that there is a general principle of information processing at the levels of both preattention and attention. It is claimed that, at both levels, information processing is based on the coherent (synchronous) activity of neurons, neural populations, and brain structures. The level difference is presumed to relate to how synchronization is realized. At the level of preattention, synchronization results from the self-organization of cortical activity, whereas at the level of attention, synchronous activity is controlled by special brain structures that act as a central executive. Two types of oscillatory neural networks are developed to model preattention and attention phenomena. In preattention modeling we concentrate on the binding problem. To solve this problem, a two-layer network of neural oscillators is developed which is able to generate two-frequency envelope oscillations, where the amplitude of high-frequency oscillations is modulated by a lower frequency. This network synchronizes regions of oscillatory activity at high and low frequencies according to the type of stimulation. Such synchronization gives feature binding for both simple and complex stimuli. Networks of phase oscillators with a central element are used to describe a different dynamical behavior that is associated with attention focus formation and switching. For input to the attention system represented by two stimuli, we give a complete description of conditions when
a specific attention focus can be formed. The results are interpreted and discussed in terms of attention modeling. This includes the interpretation of psychological experiments on visual selective attention, the problem of attention focus formation, and the possibility of spontaneous attention switching.