A Hybrid Nonlinear-Discriminant Analysis Feature Projection Technique

摘要: Feature set dimensionality reduction via Discriminant Analysis (DA) is one of the most sought after approaches in many applications. In this paper, a novel nonlinear DA technique presented based on hybrid Artificial Neural Networks (ANN) and Uncorrelated Linear (ULDA). Although ULDA can present statistically uncorrelated features, but similar to existing DA's it assumes that original data linearly separable, which not case with real world problems. order overcome problem, layer feed-forward ANN trained Differential Evolution (DE) optimization combined implement feature projection technique. This combination acts as discriminant analysis. The proposed approach validated Brain Computer Interface (BCI) problem compared other techniques.