Approximation of unknown input-output mappings by optimizing approximating functions is important for a number of practical applications. A straightforward method involves dividing the whole input space into small regions and finding the optimal approximating value within each. For such a method to work well the way the input region is divided is very important. In this paper we derive an algorithm that takes into account both the distribution of the input points and how rapidly the mapping is changing. The method is demonstrated on a simple function approximation problem.
|Number of pages||7|
|Journal||Neural Network World|
|Publication status||Published - Jan 1 1996|
ASJC Scopus subject areas
- Hardware and Architecture
- Artificial Intelligence