What differentiates Linear Discriminant Analysis (LDA) from Quadratic Discriminant Analysis (QDA)?

LDA assumes heterogeneity among class variances, meaning they each share a single covariance matrix, while QDA allows for each class to have its own variance. Thus, QDA provides additional flexibility for learning non-linear decision boundaries. It is generally recommended to try both LDA and QDA on a dataset and use cross validation to determine which performs best.