Self-Organizing Neural Networks for Multitarget Track Initiation

This paper describes our work with self-organizing neural networks which are dominated by competitive inhibition. Our research has shown that such networks will eventually cluster their internal states about the modes of a stimulating probability density function and therefore can be used in parameter estimation problems characterized by nonGaussian or multimodal densities. In particular, we present simulation results demonstrating the application of these neural networks to multitarget track initiation problems.