Analysis of feature concatenation operation on vector spaces
Mete Özay, Fatos Tunay Yarman Vural · 2010
In this study, the theoretical and experimental analysis of feature space concatenation operation is introduced. This operation is widely used for data fusion and ensemble learning. Following the analysis, a new performance measure which is called Vectorization Measure (VM) is introduced. VM enables the estimation of the separability capacity of the fusion space by analyzing the sample margin distributions on the feature vector subpsaces.