Fig. 1: Principal component analysis of the S(Q) dataset.
From: A machine learning inversion scheme for determining interaction from scattering
a The eigenvalues of principal components. SVR stands for singular value rank and the singular value is represented by Σ. As demonstrated in panel b, the results of maximum profile likelihood analysis41 show that the linear subspace spanned by the first three singular vectors given in Panel c is sufficiently expressive to illustrate the features of the correlated data points. SVD0, SVD1, and SVD2 are presented as a function of dimensionless unit QD.
