The main objective of this work is to investigate connectionist (neural network) clustering models. Several areas of focus are: 1. a new mode-seeking clustering algorithm that handles nominal attributes and small sample size; its realization as a parallel and distributed process; 2. limited lateral inhibition, particularly its effects on topographic maps; 3. artificial neurons as evidence combination units and their realization of probabilistic reasoning; 4. incremental versions of mode-seeking clustering algorithsm. The results of this work can be used in the design of new content-addressable memories, indexing methods in case-based reasoning, and other efforts in the understanding of intelligent systems.
