Outliers often reveal crucial information about the underlying data such as the presence of unusual observations that require for in-depth analysis. The detection of outliers is especially challenging in real-world application scenarios dealing with high-dimensional and flat data bearing different subpopulations of potentially varying data distributions. In the context of high-dimensional data, PCA-based methods are commonly applied in order to reduce dimensionality and to reveal outliers. In this paper, we perform a thorough empirical evaluation of well-establish PCA-based methods for the detection of outliers in a challenging audio data set. In this evaluation we focus on various experimental data settings motivated by the requirements of real-world scenarios, such as varying number of outliers, available training data, and data characteristics in terms of potential subpopulations.

Top