Estimating the spatial extent of attractors of iterated function systems.pdfVIP

Estimating the Spatial Extent of Attractors of Iterated Function Systems D. Canright Mathematics Dept., Code MA/Ca Naval Postgraduate School Monterey, CA 93943 August 1993 Abstract From any given Iterated Function System, a small set of balls that cover the fractal attractor can be simply determined. This gives a priori bounds on the region of space in which the attractor may be constructed. Introduction As Barnsley, Demko and others have shown [1, 2, 6, 4], one e?ective method for produc- ing fractal shapes (in any number of dimensions) is with Iterated Function Systems (IFSs), using the “Chaos Game” algorithm (or some deterministic algorithm). This approach has been used for producing naturalistic shapes [4], ?nding fractal interpolants to given data [6, p. 274] and fractal approximations of given functions [8], and even for visualizing arbitrary discrete sequences [7]. Indeed, any contractive IFS will give an attractor (usually of fractal dimension); thus it is possible to generate IFSs at random to explore the graphical possibil- ities, as is done in some educational software [5]. Similarly, because the attractor depends continuously on the parameters in the IFS [1], small data sets from any source could be encoded as IFSs for visualization. In implementing the IFS method, one important question is the prediction a priori of the region of space containing the fractal attractor. Without such a prediction, one could only approximately estimate the spatial extent based on calculating several points of the attractor, with no guarantee that these points are near the bounds. If as a result the portion of space represented in the computation of the attractor is too small, the result will not yield the whole attractor. If the porti