Fuzzy Lattice Neurocomputing Using Weighted Cosine Similarity Measure

In this work, we investigate the effects of changing the underlying inclusion measure used by fuzzy lattice neurocomputing (FLN) classifiers to the cosine similarity measures. We also show that by weighing the contribution of each attribute found in the data set, we can provide additional improvements over simply using an inclusion/similarity measure. Furthermore, we show via experiments using differential evolution and weighted cosine similar measures that the proposed techniques imply significant improvements.