800 likes | 1.12k Views
Rank Annihilation Based Methods. p. X. n. rank(P p ) = rank(P n ) = rank(X) < min (n, p). Rank. The rank of matrix X is equal to the number of linearly independent vectors from which all p columns of X can be constructed as their linear combination.
E N D
p X n rank(Pp) = rank(Pn) = rank(X) < min (n, p) Rank The rank of matrix X is equal to the number of linearly independent vectors from which all p columns of X can be constructed as their linear combination Geometrically, the rank of pattern of p point can be seen as the minimum number of dimension that is required to represent the p point in the pattern together with origin of space
y P Q R O xP x Variance xP yP xQ yQ xR yR OP2= xP2 + yP2 yP OQ2= xQ2 + yQ2 yQ OR2= xR2 + yR2 yR OP2 + OQ2 + OR2= xP2 + yP2 + xQ2 + yQ2 + xR2 + yR2 xQ xR Sum squared of all elements of a matrix is a criterion for variance in that matrix
Eigenvectors and Eigenvalues 0 v R I R v = l - l v For a symmetric, real matrix, R, an eigenvector v is obtained from: Rv = lv l is an unknown scalar-the eigenvalue (R – lI) v= 0 Rv – vl = 0 The vector v is orthogonal to all of the row vector of matrix (R-lI) =
Variance and Eigenvalue 1 2 D = 2 4 3 6 1 2 D = 4 2 3 6 Rank (D) = 1 Variance (D) = (1)2 + (2)2 + (3)2 + (2)2 + (4)2 + (6)2 = 70 Eigenvalues (D) = [70 0] Rank (D) = 2 Variance (D) = (1)2 + (4)2 + (3)2 + (2)2 + (2)2 + (6)2 = 70 Eigenvalues (D) = [64.4 5.6]
Free Discussion The relationship between eigenvalue and variance
10 uni-components samples 204.7 0 0 0 0 0 0 0 0 0 204.6 0.0012 0.0012 0.0011 0.0009 0.0009 0.0008 0.0006 0.0006 0.0005 Eigenvalues (D) = Eigenvalues (D) = Without noise with noise
10 bi-components samples 262.65 18.94 0 0 0 0 0 0 0 0 262.64 18.93 0.0011 0.0009 0.0008 0.0008 0.0007 0.0007 0.0006 0.0005 Eigenvalues (D) = Eigenvalues (D) = Without noise with noise
= Bilinearity As a convenient definition, the matrices of bilinear data can be written as a product of two usually much smaller matrices. D = C E + R
Bilinearity in multi-component absorbing systems k2 k1 A B C D = C ST D = cA sAT + cB sBT + cC sCT
Bilinearity cA cA sAT sAT
Bilinearity cB cB sBT sBT
Bilinearity cC cC sCT sCT
cA sAT cB sBT cC sCT + +
Spectrofluorimetric spectrum Emission wavelength EEM Excitation wavelength Excitation-Emission Matrix (EEM) is a good example of bilinear data matrix
Free Discussion Why the EEM is bilinear?
Trilinearity C=1.0 C=0.8 C=0.6 C=0.4 C=0.2
Quantitative Determination by Rank Annihilation Factor Analysis
Two components mixture of x and y Cx=1.0 and Cy=2.0