High-dimensional classification is an important and challenging statistical problem. We develop a set of quadratic discriminant rules by simplifying the structure of the covariance matrices instead of ...