sitator.util package¶

Submodules¶

sitator.util.elbow module¶

sitator.util.elbow.index_of_elbow(points)¶

Returns the index of the “elbow” in points.

Decently fast and pretty approximate. Performs worse with disproportionately long “flat” tails. For example, in a dataset with a nearly right-angle elbow, it overestimates the elbow by 1 starting at a before/after ratio of 1/4.

sitator.util.mcl module¶

sitator.util.mcl.markov_clustering(transition_matrix, expansion=2, inflation=2, pruning_threshold=1e-05, iterlimit=100)¶

Compute the Markov Clustering of a graph. See https://micans.org/mcl/.

Because we’re dealing with matrixes that are stochastic already, there’s no need to add artificial loop values.

Implementation inspired by https://github.com/GuyAllard/markov_clustering

sitator.util.progress module¶

sitator.util.zeo module¶

class sitator.util.zeo.Zeopy(path_to_zeo)¶

Bases: object

A wrapper for the Zeo++ network tool.

Warning:	Do not use a single instance of Zeopy in parallel.

static ase2cuc(at)¶

Convert an ase.Atoms to the CUC format.

See http://www.maciejharanczyk.info/Zeopp/input.html

Returns:	A string in CUC format.

static parse_nt2(nt2lines)¶

static parse_v1_cell(v1lines)¶

voronoi(structure, radial=False)¶

Parameters:	structure (Atoms) – The ASE Atoms to compute the Voronoi decomposition of.

Module contents¶

class sitator.util.DotProdClassifier¶

Bases: object

Assign vectors to clusters indicated by a representative vector using a cosine metric.

Cluster centers can be given through set_cluster_centers() or approximated using the custom method described in the appendix of the main landmark analysis paper (fit_centers()).

Parameters:

threshold (float) – Similarity threshold for joining a cluster. In cos-of-angle-between-vectors (i.e. 1 is exactly the same, 0 is orthogonal)
max_converge_iters (int) – Maximum number of iterations. If the algorithm hasn’t converged by then, it will exit with a warning.
min_samples (float or int) – filter out clusters with low sample counts. If an int, filters out clusters with fewer samples than this. If a float, filters out clusters with fewer than floor(min_samples * n_assigned_samples) samples assigned to them.

cluster_centers¶

cluster_counts¶

fit_centers¶

fit_predict¶: Fit the data vectors X and return their cluster labels.

n_clusters¶

predict¶

Return a predicted cluster label for vectors X.

Parameters:	threshold (float) – alternate threshold. Defaults to None, when self.threshold is used.
Returns:	an array of labels. -1 indicates no assignment.
Returns:	an array of confidences in assignments. Normalzied values from 0 (no confidence, no label) to 1 (identical to cluster center).

set_cluster_centers¶

class sitator.util.PBCCalculator¶

Bases: object

Performs calculations on collections of 3D points under PBC.

all_in_unit_cell()¶

average()¶

Average position of a “cloud” of points using the shift-and-wrap hack.

Copies the points.

Assumes that the points are relatively close (within a half unit cell) together, and that the first point is not a particular outsider (the cell is centered at that point). If the average is weighted, the maximally weighted point will be taken as the center.

Can be a weighted average with the semantics of :func:numpy.average.

cell_centroid¶

distances()¶

Compute the Euclidean distances from pt1 to all points in pts2, using shift-and-wrap.

Makes a copy of pts2 unless in_place == True.

Returns ndarray len(pts2):
	distances

is_in_image_of_cell()¶

is_in_unit_cell()¶

min_image()¶

Find the minimum image of pt relative to ref. In place in pt.

Uses the brute force algorithm for correctness; returns the minimum image.

Assumes that ref and pt are already in the same cell (though not necessarily the <0,0,0> cell – any periodic image will do).

Returns int[3] minimg:
	Which image was the minimum image.

pairwise_distances()¶

Compute the pairwise distance matrix of pts with itself.

Returns ndarray (len(pts), len(pts)):
	distances

time_average()¶

Do multiple PBC correct means. Frames is n_frames x n_pts x 3.

Returns a time average the size of one frame.

to_cell_coords()¶: Convert to cell coordinates in place.

to_real_coords()¶: Convert to real coords from crystal coords in place.

wrap_point()¶: Wrap a single point into the unit cell, IN PLACE. 3D only.

wrap_points()¶: Wrap points into a unit cell, IN PLACE. 3D only.

class sitator.util.RecenterTrajectory¶

Bases: object

run¶

Recenter a trajectory.

Recenters traj on the center of mass of the atoms indicated by static_mask, IN PLACE.

Parameters:

structure (ase.Atoms) – An atoms representing the structure of the simulation.
static_mask (ndarray) – Boolean mask indicating which atoms to recenter on
positions (ndarray) – (n_frames, n_atoms, 3), modified in place
velocities (ndarray, optional) – Same; modified in place if provided
masses (None or dict or ndarray) –
The masses to use when computing the center of mass.
- If None, masses from structure.get_masses() will
  
  be used.
- If a dict, expected to map chemical symbols to masses
- If an ndarray, must have n_atoms elements giving the
  
  masses of all atoms in the system.