Neural networks have become a popular tool for function approximation and classification. Table 1 summarizes some of the tradeoffs between using them and using memory based learning. Some of these issues are described in further detail below, while others are addressed later in the tutorial.