11.07.2015 Views

2DkcTXceO

2DkcTXceO

2DkcTXceO

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

L.A. Wasserman 531●●●●●●●●●●●●●●● ●●●●●●● ●●●●● ●● ●●●●●●●●●FIGURE 44.3The mean shift algorithm. The data points move along trajectories duringiterations until they reach the two modes marked by the two large asterisks.The mean shift algorithm (Fukunaga and Hostetler, 1974; Comaniciu andMeer, 2002) is a method for finding the modes of a density by following thesteepest ascent paths. The algorithm starts with a mesh of points and thenmoves the points along gradient ascent trajectories towards local maxima. Asimple example is shown in Figure 44.3.Given a function p : R D → R, letg(x) =∇p(x) denotethegradientatxand let H(x) denotetheHessianmatrix.Letλ 1 (x) ≥···≥λ D (x) (44.8)denote the eigenvalues of H(x) and let Λ(x) bethediagonalmatrixwhosediagonal elements are the eigenvalues. Write the spectral decomposition ofH(x) as H(x) =U(x)Λ(x)U(x) ⊤ .Fix0≤ d

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!