查看文章

Detailed measurements were performed in two real life medical domains in order to verify the importance of redundant knowledge. We compared performance of the GINESYS system with and without redundant knowledge and varied the number of learning examples, the amount of artificially added noise, the impurity and the error estimate functions. Our attempts to make a theoretical model explaining the mechanism of redundant knowledge are still unsuccessful but may be helpful for further research. It seems that one must abandon general domain independent model and capture the specific nature of real life domains into the model. Finally, measurements confirm that redundant knowledge is essential for superior behavior of inductive learning systems and they also strongly indicate that redundant knowledge is a superior knowledge representation over'classical'representations in real life noisy and incomplete domains.