In this problem we explore the "leave-one-out method," discussed in Chapter 4, for estimating the expected test error produced by an optimal hyperplane for the case of nonseparable patterns. Discuss the various possibilities that can arise in the use of this method by eliminating any one pattern from the training sample and constructing a solution based on the remaining patterns.