Visualisation of artificial neural networks to assist in application development
Gary Whittington, C.T. Spracklen · 1991
The importance of visualisation of scientific data has increased over recent years and has had a diverse range of applications. However, within the field of artificial neural networks (ANN), visualisation has been limited to comparatively simple techniques. This is especially surprising considering the strong geometric and physical analogies present within the ANN field. The paper examines the potential for various visualisation techniques in the design and synthesis of ANN's. Descriptions of various visualisation techniques are drawn from the authors' research area, the adaptive Kohonen feature map model, and from practical design processes associated with the development of tracking and classification systems. The paper is divided into three sections: a brief introduction to scientific visualisation, visualisation of ANN's, and applying visualisation techniques as an aid to ANN design.