Distance weighted discrimination

High Dimension Low Sample Size statistical analysis is becoming in-creasingly important in a wide range of applied contexts. In such sit-uations, it is seen that the appealing discrimination method called the Support Vector Machine can be improved. The revealing concept is data piling at the margin. This leads naturally to the development of Distance Weighted Discrimination, which also is based on modern com-putationally intensive optimization methods, and seems to give improved generalizability.