Comparing learning accuracies of neural nets and decision-tree classifier systems

A typical neural net algorithm and a typical decision-tree classifier are described. Although the two strategies approach learning differently, research suggests that the methods may be used to complement each other. One major difficulty in this regard is comparing learning accuracies of neural net algorithms and decision-tree classifier systems. A description is given of the learning accuracy problem and research efforts which are expected to achieve a solution are outlined. Numerical experiments that illustrate the accuracy problem are included.<<ETX>>