A convex hull algorithm for neural networks

A convex hull algorithm for neural networks is presented. It is applicable in both two and three dimensions and has a time complexity of O(N) for the offline case, O(log N) for the online case in two dimensions, and O(hN), O(N), respectively, for three dimensions (h is the number of faces in the convex hull). The constant bounding the complexity is expected to be very small.<
>