Outliers detection in ICA
Pingxing Feng, Liping Li, Hongbo Zhang, Guobin Qian · 2013
Outliers have a significant influence on separate performance of independent component analysis (ICA). Unfortunately, the traditional methods used in ICA do not consider the influence of outliers. In this work an influence function-based detection method is introduced to find the outliers in ICA. Traditional outliers detection techniques can not be directly applied to ICA due to the nature of non-cooperate observed data and limitations of the independent components. This work provides a influence function-based technique to find the outliers in the observed signals. Simulations results show the effectiveness of the proposed approach to detect and establish the outliers in the observed signal.