Mind in Motion

The utilization of noise in the cognitive processs

Robert Pallbo © 1997

Ph.D. Thesis, Lund University Cognitive Studies 57,
ISBN 91-628-2708-1
.

Robert Pallbo » Academic » Papers » Mind in Motion

Content
Abstract

The brain is not silent. Even in the absence of stimuli are the neurons activated every now and then in what is called ``spontaneous'' or ``background activity.'' This activity provides a noisy background to the operation of the brain. This work shows how this noise can be utilized in the cognitive process. That is, rather than being treated as a nuisance, noise is given a functional role in the brain's activity. More than this, the role noise is given is a very essential one. In fact, the models presented in this work would not function without it.

The usefulness of the spontaneous activity is demonstrated in two ways. First, it is shown how this activity can enable new knowledge to be assimilated by the system. This ability is described using a framework of generalised evolution. The framework permits the cognitive process to be considered as an evolution occurring within the life-time of the individual. In this evolutionary view of cognition, the spontaneous activity provides the source of variation. This source of variation is very simple in its nature. Yet its effects are able to follow the growth of an ever more complex cognitive apparatus as it evolves in the individual mind. The reason for this ability is that the spontaneous activity operates directly on the cognitive structures, i.e., the neurons and their connections, and does therefore not need any interpretation as a genetic mechanism would.

The second way in which the usefulness of noise is demonstrated is in the form of a model of how the brain is able to detect visual motion. In the brain, there are neurons that are specialised to detect motion of a certain direction in the visual field of the eyes. The model presented here is intended as a model of how these neurons operate. Furthermore, as this model operates in the same way as the assimilation of knowledge mentioned above, it serves as a concrete example of an application of those ideas.

In the field of motion detection, the model operates in a very novel way. First of all, it makes the basic assumption that the motion present in the current scene was already detected in the previous scene. That is, all that the model needs to do is to update this detection to fit the current scene. This reduces the complexity of the model considerably as no initial motion needs to be detected. Instead, initial motion is captured by the system by utilizing the spontaneous activity. Any motion in the visual scene is assimilated by the process if it harmonizes with the background activity. Thus, the operation of the system is dependent on the presence of noise. Just like the case of knowledge acquisition mentioned above, noise is used to assimilate new structures into the system.

Copyright © 1995-2005, Robert Pallbo