Cyberspace geography visualization
Using the topology of the World-Wide Web, a metric as been defined. For this purpose, modelling through graphs has been made and the distances between resources were defined as the shortest path in the graph. This enabled the possibility of representing each resource as a point in a high-dimensional space.
To permit a display of the interactions of the resources placed in a space of high dimensionality, a transformation has been required. This transformation has been made by defining a mapping of the resources onto low-dimensional visualization media, typically lattices with two dimensions. The constraint made on the mapping was to keep the monotonicity. Thus, only the rank order of the distances are preserved, protecting the most important features of the system.
To perform this non-linear dimensionality reduction, the self-organizing maps algorithm has been shown to provide the best results. The self-organizing maps method is an unsupervised neural network model that produce topology preserving maps.
Although, a model for topologically-organized maps was not sufficient because the order of similarity between resources cannot be visualized. To address with this problem, a model for representing the reliefs of self-organizing maps, the unified matrix method, was followed. By analysing the weight vectors of the self-organizing maps, a representation of the landscape became possible. It was then possible to interpret the components of this landscape as being the equivalent to mountains, ravines and valleys.
Based upon the above mentioned methods, an experimental prototype has been built. This included a software agent to gather the information, a program to immerse resources in a high dimensional space, the computation of the self-organizing maps to produce maps of low dimensionality, the creation of the unified matrix to represent the reliefs, and the development of a graphical user interface to permit the visualization of the resulting geographical maps and to give the possibility to access directly the resources behind the maps.
Since the prototype was only of academic purpose, various improvements were sketched to improve its usability. The methods used were made scalable in their spirit and therefore taking scalability into account was also possible.
The results, made available in the World-Wide Web, were shown to provide an original way to improve cyberspace navigability and to address the lost-in-cyberspace syndrome problem. It is thus encouraged that further research be done in this direction.