Software Tool For Visualization Of Neural Network Gaps

This paper presents a software tool for visualization of neural network gaps. These gaps represent significant differences between the intended conceptualization, and the trained generalization obtained by an ANN. We tested our software on two-real world applications. The first is in applying neural networks to the diagnostics of Polycythemia Vera; the second is based on 1994 Census database with more than 40,000 records.