ABSTRACT
This chapter begins by deriving the Quadratic Discriminant analysis as the boundary between two normally distributed classes. The MATLAB code applies the Quadratic Discriminant Analysis to Fisher’s iris data. The kernel trick can be applied to linear classification methods to separate data which is not linearly separable. The kernel function was derived from a higher dimensional inner product. The hyperbolic tangent kernel addresses the problem that absolute values of the inner product increase with the distance from the zero line. Neural networks are explored further as a set of many conditions of the form to the left or right of a line determined by the network parameters. Neural networks are dynamic systems characterized by non-linear, distributed, parallel and local processing. A neural network consists of neurons, also known as nodes node or units and synapses connecting the neurons. The chapter concludes with boosting and cascades as a way to build strong classifiers from weak ones and taking the cost into account.
