Where fj is the j-th feature from a set ofFfeatures and the ci are the C principal components. The variance of the i-th PC is denoted by σ2(ci). The factor loading matrix L is a RCxF matrix with C columns and F rows.
The WALDI of a feature is a measure of its information content in terms of orthogonal variances in the data. This value is proportional to the expressiveness of a feature. However, it does not contain information about redundancies among different features. This information can be derived from the factor loading matrix.