On Optimal Projective Fusers for Function Estimators
Nageswara S. V. Rao · University of North Texas Digital Library (University of North Texas) · 1999
We propose a fuser that projects different function estimators in different regions of the input space based on the lower envelope of the error curves of the individual estimators. This fuser is shown to be optimal among projective fusers and also to perform at least as well as the best individual estimator. By incorporating an optimal linear fuser as another estimator, this fuser performs at least as well as the optimal linear combination. We illustrate the fuser by combining neural networks trained using different parameters for the network and/or for learning algorithms. Keywords: Sensor fusion, fusion rule estimation, empirical estimation 1 Introduction Recently, combinations of estimators have been shown to be very effective in a number of disciplines such as forecasting, reliability, and pattern recognition (see [6] for an overview). In specific methods such as neural networks, it has been shown that better performance can be achieved by suitably combining the networks rather th...