qstack.mathutils.fps

Farthest Point Sampling algorithm implementation.

qstack.mathutils.fps.do_fps(x, d=0)[source]

Perform Farthest Point Sampling on a set of points.

References

P. O. Dral, A. Owens, S. N. Yurchenko, W. Thiel, “Structure-based sampling and self-correcting machine learning for accurate calculations of potential energy surfaces and vibrational levels”, J. Chem. Phys. 146 244108 (2017), doi:10.1063/1.4989536

G. Imbalzano, A. Anelli, D. Giofré, S. Klees, J. Behler, M. Ceriotti, “Automatic selection of atomic fingerprints and reference configurations for machine-learning potentials”, J. Chem. Phys. 148 241730 (2018), doi:10.1063/1.5024611

K. Rossi, V. Jurásková, R. Wischert, L. Garel, C. Corminboeuf, M. Ceriotti, “Simulating solvation and acidity in complex mixtures with first-principles accuracy: the case of CH3SO3H and H2O2 in phenol”, J. Chem. Theory Comput. 16 5139–5149 (2020), doi:10.1021/acs.jctc.0c00362

Code from Giulio Imbalzano.

Parameters:

  • x (numpy.ndarray) – 2D array of points, shape (n_points, n_features).

  • d (int) – Number of points to sample. If 0, samples all points. Defaults to 0.

Returns:

A tuple containing: - iy (numpy.ndarray): Indices of sampled points. - measure (numpy.ndarray): Distances to nearest selected point for each iteration.

Return type:

tuple