Almost Linear Biobasis Function Neural Networks
Liwen You, Thorsteinn Rögnvaldsson · IEEE International Conference on Neural Networks/IEEE ... International Conference on Neural Networks · 2007
An analysis of biobasis function neural networks is presented, which shows that the similarity metric used is a linear function and that bio-basis function neural networks therefore often end up being just linear classifiers in high dimensional spaces. This is a consequence of four things: the linearity of the distance measure, the normalization of the distance measure, the recommended default values of the parameters, and that biological data sets are sparse.