Simplex Volume Maximization



Given a collection of data points, Simplex Volume Maximization algorithms attempt to select data points that are far from each other. This is useful in data analysis. Here, the SiVM algorithm selects 4 prototypes out of 200 two dimensional Gaussian data points. This happens in 2 stages: 1. stage [0:00 to 0:32]: using the FastMap heuristic, the algorithm looks for 2 points that are particularly far apart, i.e. it selects a random point x1, looks for a point x2 that is far from x1; once x2 has been determined the algorithm searches for a point x3 far from x2; once x3 has been found, a point x4 is searched for that is far from x3; at the end of this stage, x3 and x4 are selected as the first two prototypes 2. stage [0:33 to 0:53]: given x3 and x4, the algorithm searches for a point x5 that far from both initial prototypes; once x5 is available, the final prototype x6 is determined in a similar fashion

منتشر شده توسط: bitLectures
تاریخ انتشار: ۷ سال پیش
دسته بندی: آموزشی