A NOVEL MAP PROJECTION USING AN ARTIFICIAL NEURAL NETWORK

Cartographers have become increasingly interested in an area of scientific investigation known as geocomputation (1, 2). Most of this cartographic involvement has focused on geographic visualization (3), drawing on centuries of accumulated cartographic expertise, combined with lessons learned from a diverse group of other fields, including cognitive science and computer science. What most of these activities have in common is the goal of discovering/conveying knowledge contained in increasingly large geographic data repositories. On the other hand, there are cartographic activities aimed at the metaphorical transfer of cartographic principles and methods to non-geographic domains. This is on one hand driven by the desire to utilize the spatio-cognitive abilities of humans towards data mining and knowledge discovery in non-spatial databases. On the other hand, a cartographic interpretation of novel computational techniques may help to inform and critically engage those who hope to employ these methods in non-geographic visualization. This is what this paper attempts, with a view of the self-organizing map method. It engages the distortions caused by this popular artificial neural network approach in two ways. First, the paper presents a cartographically informed method to visualize these distortions. Second, it examines whether distortion characteristics could be exploited towards a novel type of map projection.