Interpretation of models induced by artificial neural networks is often a difficult task. In this paper we focus on a relatively novel neural network architecture and learning algorithm, bp-som that offers possibilities to overcome this difficulty. It is shown that networks trained with BP-SOM show interesting regularities, in that hidden-unit activations become restricted to discrete values, and that the som part can be exploited for automatic rule extraction.
|Title of host publication||Machine learning : ECML-98 : 10th European conference on machine learning, Chemnitz, Germany, April 21-23, 1998 : proceedings|
|Editors||C. Nédellec, C. Rouveirol|
|Place of Publication||Berlin|
|Publication status||Published - 1998|
|Name||Lecture notes in artificial intelligence|
Weijters, A. J. M. M., Bosch, van den, A. P. J., & Herik, van den, H. J. (1998). Inelligible neural networks with BP-SOM. In C. Nédellec, & C. Rouveirol (Eds.), Machine learning : ECML-98 : 10th European conference on machine learning, Chemnitz, Germany, April 21-23, 1998 : proceedings (pp. 406-411). (Lecture notes in artificial intelligence; Vol. 1398). Berlin: Springer. https://doi.org/10.1007/BFb0026711