Detailed Description

This module provides methods for clustering, histograms and other statistical computations.

This module provides code to compute clusterings. Adaptors are provided that allow easy clustering of points, and configurations of models in IMP::ConfigurationSet objects among other things.

Examples:

Author(s): Keren Lasker, Daniel Russel

Version: SVN.r12662

License: LGPL. This library is free software; you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version.

Publications:

Daniel Russel, Keren Lasker, Ben Webb, Dina Schneidman, Javier Velázquez-Muriel, Andrej Sali, “Putting the pieces together: integrative structure determination of macromolecular assemblies”, PLoS Biology, 2012. This paper provides an overview of the key concepts in IMP and how to apply them to biological problems.
Frank Alber, Friedrich Foerster, Dmitry Korkin, Maya Topf, Andrej Sali, “Integrating diverse data for structure determination of macromolecular assemblies”, Annual Review of Biochemistry, 2008. This paper provides a review of the integrative structure determination methodology and various data sources that can be used.

Classes
class	ConfigurationSetRMSDMetric
class	ConfigurationSetXYZEmbedding
	Embed a configuration using the XYZ coordinates of a set of particles. More...
class	Embedding
	Map clustering data to spatial positions. More...
class	HistogramD
class	Metric
	Compute a distance between two elements to be clustered. More...
class	ParticleEmbedding
class	PartitionalClustering
	The base class for clusterings of data sets. More...
class	PartitionalClusteringWithCenter
class	RecursivePartitionalClusteringEmbedding
class	RecursivePartitionalClusteringMetric
class	VectorDEmbedding
	Simply return the coordinates of a VectorD. More...
Python only
This functionality is only available in python.
void	show_histogram (HistogramD h, std::string xscale="linear", std::string yscale="linear", Functions curves=Functions())
Functions
PartitionalClusteringWithCenter *	create_bin_based_clustering (Embedding *embed, double side)
PartitionalClustering *	create_centrality_clustering (Metric *d, double far, int k)
PartitionalClustering *	create_centrality_clustering (Embedding *d, double far, int k)
PartitionalClustering *	create_connectivity_clustering (Metric *metric, double dist)
PartitionalClusteringWithCenter *	create_connectivity_clustering (Embedding *embed, double dist)
PartitionalClustering *	create_diameter_clustering (Metric *d, double maximum_diameter)
PartitionalClusteringWithCenter *	create_lloyds_kmeans (Embedding *embedding, unsigned int k, unsigned int iterations)
algebra::VectorKDs	get_centroids (Embedding d, PartitionalClustering pc)
Ints	get_representatives (Embedding d, PartitionalClustering pc)
void	validate_partitional_clustering (PartitionalClustering *pc, unsigned int n)

Function Documentation

PartitionalClusteringWithCenter* IMP::statistics::create_bin_based_clustering	(	Embedding *	embed,
		double	side
	)

The space is grided with bins of side size and all points that fall in the same grid bin are made part of the same cluster.

PartitionalClustering* IMP::statistics::create_centrality_clustering	(	Metric *	d,
		double	far,
		int	k
	)

Cluster by repeatedly removing edges which have lots of shortest paths passing through them. The process is terminated when there are a set number of connected components. Other termination criteria can be added if someone proposes them.

Only items closer than far are connected.

PartitionalClustering* IMP::statistics::create_centrality_clustering	(	Embedding *	d,
		double	far,
		int	k
	)

Cluster by repeatedly removing edges which have lots of shortest paths passing through them. The process is terminated when there are a set number of connected components. Other termination criteria can be added if someone proposes them.

PartitionalClustering* IMP::statistics::create_connectivity_clustering	(	Metric *	metric,
		double	dist
	)

Two points, $p_i$ , $p_j$ are in the same cluster if there is a sequence of points $\left(p^{ij}_{0}\dots p^{ij}_k\right)$ such that $\forall l ||p^{ij}_l-p^{ij}_{l+1}|| < d$ .

PartitionalClusteringWithCenter* IMP::statistics::create_connectivity_clustering	(	Embedding *	embed,
		double	dist
	)

Two points, $p_i$ , $p_j$ are in the same cluster if there is a sequence of points $\left(p^{ij}_{0}\dots p^{ij}_k\right)$ such that $\forall l ||p^{ij}_l-p^{ij}_{l+1}|| < d$ .

PartitionalClustering* IMP::statistics::create_diameter_clustering	(	Metric *	d,
		double	maximum_diameter
	)

Cluster the elements into clusters with at most the specified diameter.

IMP::statistics::create_lloyds_kmeans	(	Embedding *	embedding,
		unsigned int	k,
		unsigned int	iterations
	)

Return a k-means clustering of all points contained in the embedding (ie [0... embedding->get_number_of_embeddings())). These points are then clustered into k clusters. More iterations takes longer but produces a better clustering.

The algorithm uses algebra::EuclideanVectorKDMetric for computing distances between embeddings and cluster centers. This can be parameterized if desired.

Examples: basic optimization, kmeans, nup84 cg, nup84 rb, analyze 0

algebra::VectorKDs IMP::statistics::get_centroids	(	Embedding *	d,
		PartitionalClustering *	pc
	)

Given a clustering and an embedding, compute the centroid for each cluster

Ints IMP::statistics::get_representatives	(	Embedding *	d,
		PartitionalClustering *	pc
	)

Given a clustering and an embedding, compute a representatative element for each cluster.

void show_histogram	(	HistogramD	h,
		std::string	xscale = `"linear"`,
		std::string	yscale = `"linear"`,
		Functions	curves = `Functions()`
	)

In python, you can use matplot lib, if installed, to show the contents of a histogram. At the moment, only 1D and 2D histograms are supported.

Parameters:

[in]	h	The histogram to show, the plot is sized to the histograms bounding box.
[in]	xscale	Whether the xscale is "linear" or "log"
[in]	yscale	Whether the yscale is "linear" or "log"
[in]	curves	A list of python functions to plot on the histogram as curves. The functions should take one float and return a float.

void IMP::statistics::validate_partitional_clustering	(	PartitionalClustering *	pc,
		unsigned int	n
	)

Check that the clustering is a valid clustering of n elements. An exception is thrown if it is not, if the build is not a fast build.

Detailed Description

Classes

Python only

Functions

Function Documentation