You observe that PCA explains a majority of the variance wit…
You observe that PCA explains a majority of the variance with the first several components, while ICA shows the highest excess kurtosis in its first several components. However, ICA also identifies 2 additional components with mild excess kurtosis (0.3 and 0.1). Suppose your downstream task is unsupervised clustering of eel behavioral states using these reduced representations. Which of the following best explains why choosing more than 3 components (e.g., 5) might degrade clustering performance? Choose all that apply.
Read Details