lec05非参数技术密度估计窗近邻居指标分类我们现在对似然性.pdfVIP

Nonparametric Techniques ✓ Density Estimation ✓ Parzen Windows ✓Nearest-Neighbor ✓ Metrics and Nearest-Neighbor Classification Density Estimation We now make NO assumptions about the form of the probability density function of the likelihood p( | x). j There are two types of nonparametric methods: Estimating p(x |  ) j Bypass probability and go directly to a-posteriori probability estimation Density Estimation To beg onsider a point x in the feature space. Assume x belongs to a region R. The probability P of that region is as follows: P p (x )dx (1)  for all x’in region R Prob. Of Region R Feature Space R x P = Prob( R ) Question: how can we estimate p(x) given that we know P? Example R : { x | 20 cm length(x) 30 cm } What is the probability of x = 25cm ? Estimation of p(x) P is a smoothed (or averaged) version of the density function p(x). If we have a sample of size n, therefore, the probability that k points fall in R is then: n   k n−k Pk  P ( 1 −P ) (2) k   and the expected value for k is: E(k) = nP (3) Estimation of p(x) ML estimation of P =  ˆ k Max( Pk | ) is reached for   P  n Therefore, the ratio k/n is a good estimate for the probability P and hence for the density function p. Estimation of p(x) p(x) is continuous and that the region