Search

It is generally believed that cognition involves the self-organization of coherent dy- namic functional networks across several brain regions in response to incoming stimulus and internal modulation. These context-dependent networks arise continually from the spatiotemporally multi-scale structural substrate of the brain configured by evolution, development and previous experience, persisting for 100–200 ms and generating re- sponses such as imagery, recall and motor action. In the current paper, we show that a system of interacting modular attractor networks can use a selective mechanism for assembling functional networks from the modular substrate. We use the approach to develop a model of idea-generation in the brain. Ideas are modeled as combinations of concepts organized in a recurrent network that reflects previous associations between them. The dynamics of this network, resulting in the transient co-activation of concept groups, is seen as a search through the space of ideas, and attractor dynamics is used to “shape” this search. The process is required to encompass both rapid retrieval of old ideas in familiar contexts and efficient search for novel ones in unfamiliar situations (or during brainstorming). The inclusion of an adaptive modulatory mechanism allows the network to balance the competing requirements of exploiting previous learning and exploring new possibilities as needed in different contexts.

Randomly deployed wireless sensor networks are becoming increasingly viable for many applications. Such networks can comprise anywhere from a few hundred to thousands of sensor nodes, and these sizes are likely to grow with advancing technology, making scalability a primary concern. Each node in these sensor networks is a small unit with limited resources and localized sensing and communication. Thus, all global tasks must be accomplished through self-organized distributed algorithms, which also leads to improved scalability, robustness and flexibility. In this paper, we examine the use of distributed algorithms to infer the spatial structure of an extended environment monitored by a self-organizing sensor network. Based on its sensing, the network segments the environment into regions with distinct characteristics, thereby inferring a cognitive map of the environment. This, in turn, can be used to answer global queries about the environment efficiently and accurately. The main challenge to the network arises from the necessarily irregular spatial sampling and the need for totally distributed computation. We consider distributed machine learning techniques for segmentation and study the variation of segmentation quality with reconstruction at different node densities and in environments of varying complexity.

Context is an essential part of all cognitive function. However, neural network models have only considered this issue in limited ways, focusing primarily on the conditioning of a system’s response by its recent history. This type of context, which we term Type I, is clearly relevant in many situations, but in other cases, the system’s response for an extended period must be conditioned by stimuli encountered at a specific earlier time. For example, the decision to turn left or right at an intersection point in a navigation task depends on the goal set at the beginning of the task. We term this type of context, which sets the “frame of reference” for an entire episode, Type II context. The prefrontal cortex in mammals has been hypothesized to perform this function, but it has been difficult to incorporate this into neural network models. In the present chapter, we describe an approach called latent attractors that allows self-organizing neural systems to simultaneously incorporate both Type I and Type II context dependency. We demonstrate this by applying the approach to a series of problems requiring one or both types of context. We also argue that the latent attractor approach is a general and flexible method for incorporating multi-scale temporal dependence into neural systems, and possibly other self-organized