NearestNeighborGraph

NearestNeighborGraph[{elem1,elem2,}]

gives a graph with vertices elem1,elem2, and edges connecting each elemi to its nearest neighbors.

NearestNeighborGraph[{elem1,elem2,},k]

gives a graph connecting each elemi to its k nearest neighbors.

NearestNeighborGraph[{elem1,elem2,},{k,r}]

gives a graph connecting each elemi to at most k nearest vertices within radius r of elemi.

NearestNeighborGraph[{elem1,elem2,},{All,r}]

gives a graph connecting each elemi to all vertices within radius r of elemi.

Details and Options

  • The nearest neighbor graph is also known as k-nearest neighbor graph (k-NNG).
  • An element elemj is a nearest neighbor of an element elemi whenever the distance from elemi to elemj is no larger than the distance from elemi to any other element.
  • An element elemj is a k-nearest neighbor of an element elemi whenever the distance from elemi to elemj is among the k^(th) smallest distances from elemi to any other element.
  • NearestNeighborGraph works for a variety of data, including numerical, geospatial, textual, and visual.
  • NearestNeighborGraph has the same options as Graph, with the following additions and changes:
  • DirectedEdgesAutomaticwhether edges should be taken to be directed
    DistanceFunctionAutomaticthe distance metric or dissimilarity to use
    WorkingPrecisionAutomaticprecision to use for numeric data
  • NearestNeighborGraph takes the same DistanceFunction as Nearest.

Examples

open allclose all

Basic Examples  (2)

A nearest neighbor graph:

Generate the k-nearest neighbor graph for integers in the range 1 to 10:

Scope  (10)

Basic Uses  (4)

A nearest neighbor graph for integers in the range 1 to 10:

Generate the 2-nearest neighbor graph:

Generate the at most 3-nearest neighbor graph within radius 2:

Generate the nearest neighbor graph within radius 3:

Numerical Data  (1)

A nearest neighbor graph for two-dimensional numerical data:

Three-dimensional dataset:

Boolean Data  (1)

A nearest neighbor graph for a set of Boolean data:

Geo Positions  (1)

Generate the nearest geodesic distance neighbor graph:

Strings  (1)

Generate the 2-nearest strings neighbor graph:

Images  (1)

Generate the 2-nearest images neighbor graph:

Colors  (1)

Generate the 2-nearest colors neighbor graph:

Options  (89)

DistanceFunction  (7)

By default, EuclideanDistance is used for numerical data:

EditDistance is used for strings:

GeoDistance is used for geo positions:

JaccardDissimilarity is used for Boolean data:

ColorDistance is used for colors:

ImageDistance is used for images:

The DistanceFunction can be given as a symbol:

Or as a pure function:

Method  (1)

Compare different methods for machine-precision data:

In three dimensions, the "KDtree" method is faster:

In 20 dimensions, a simple scan is faster:

WorkingPrecision  (1)

Using WorkingPrecision->MachinePrecision ensures the fastest evaluation method is used:

AnnotationRules  (2)

Specify an annotation for vertices:

Edges:

DirectedEdges  (1)

By default, an undirected graph is generated:

Use DirectedEdges->True to generate a directed graph:

EdgeLabels  (7)

Label the edge 12:

Label all edges individually:

Use any expression as a label:

Use Placed with symbolic locations to control label placement along an edge:

Use explicit coordinates to place labels:

Vary positions within the label:

Place multiple labels:

Use automatic labeling by values through Tooltip and StatusArea:

EdgeShapeFunction  (6)

Get a list of built-in settings for EdgeShapeFunction:

Undirected edges including the basic line:

Lines with different glyphs on the edges:

Directed edges including solid arrows:

Line arrows:

Open arrows:

Specify an edge function for an individual edge:

Combine with a different default edge function:

Draw edges by running a program:

EdgeShapeFunction can be combined with EdgeStyle:

EdgeShapeFunction has higher priority than EdgeStyle:

EdgeStyle  (2)

Style all edges:

Style individual edges:

EdgeWeight  (2)

Specify a weight for all edges:

Use any numeric expression as a weight:

GraphHighlight  (3)

Highlight the vertex 1:

Highlight the edge 23:

Highlight vertices and edges:

GraphHighlightStyle  (2)

Get a list of built-in settings for GraphHighlightStyle:

Use built-in settings for GraphHighlightStyle:

GraphLayout  (5)

By default, the layout is chosen automatically:

Specify layouts on special curves:

Specify layouts that satisfy optimality criteria:

VertexCoordinates overrides GraphLayout coordinates:

Use AbsoluteOptions to extract VertexCoordinates computed using a layout algorithm:

PlotTheme  (4)

Base Themes  (2)

Use a common base theme:

Use a monochrome theme:

Feature Themes  (2)

Use a large graph theme:

Use a classic diagram theme:

VertexCoordinates  (3)

By default, any vertex coordinates are computed automatically:

Extract the resulting vertex coordinates using AbsoluteOptions:

Specify a layout function along an ellipse:

Use it to generate vertex coordinates for a graph:

VertexCoordinates has higher priority than GraphLayout:

VertexLabels  (13)

Use vertex names as labels:

Label individual vertices:

Label all vertices:

Use any expression as a label:

Use Placed with symbolic locations to control label placement, including outside positions:

Symbolic outside corner positions:

Symbolic inside positions:

Symbolic inside corner positions:

Use explicit coordinates to place the center of labels:

Place all labels at the upper-right corner of the vertex and vary the coordinates within the label:

Place multiple labels:

Any number of labels can be used:

Use the argument Placed to control formatting including Tooltip:

Or StatusArea:

Use more elaborate formatting functions:

VertexShape  (5)

Use any Graphics, Image, or Graphics3D as a vertex shape:

Specify vertex shapes for individual vertices:

VertexShape can be combined with VertexSize:

VertexShape is not affected by VertexStyle:

VertexShapeFunction has higher priority than VertexShape:

VertexShapeFunction  (10)

Get a list of built-in collections for VertexShapeFunction:

Use built-in settings for VertexShapeFunction in the "Basic" collection:

Simple basic shapes:

Common basic shapes:

Use built-in settings for VertexShapeFunction in the "Rounded" collection:

Use built-in settings for VertexShapeFunction in the "Concave" collection:

Draw individual vertices:

Combine with a default vertex function:

Draw vertices using a predefined graphic:

Draw vertices by running a program:

VertexShapeFunction can be combined with VertexStyle:

VertexShapeFunction has higher priority than VertexStyle:

VertexShapeFunction has higher priority than VertexSize:

VertexShapeFunction has higher priority than VertexShape:

VertexSize  (8)

By default, the size of vertices is computed automatically:

Specify the size of all vertices using symbolic vertex size:

Use a fraction of the minimum distance between vertex coordinates:

Use a fraction of the overall diagonal for all vertex coordinates:

Specify size in both the and directions:

Specify the size for individual vertices:

VertexSize can be combined with VertexShapeFunction:

VertexSize can be combined with VertexShape:

VertexStyle  (5)

Style all vertices:

Style individual vertices:

VertexShapeFunction can be combined with VertexStyle:

VertexShapeFunction has higher priority than VertexStyle:

VertexStyle can be combined with BaseStyle:

VertexStyle has higher priority than BaseStyle:

VertexShape is not affected by VertexStyle:

VertexWeight  (2)

Set the weight for all vertices:

Use any numeric expression as a weight:

Applications  (7)

Basic Uses  (2)

Fixed-radius near neighbors:

-approximate nearest neighbor search:

Graphs & Networks  (2)

Farthest neighbor graphs:

Generate a spatial random graph within radius 0.2:

Geometry  (1)

Simplification of point-sample surfaces:

Proximity Searching  (2)

Search nearest image neighbors:

Search the dictionary neighbors:

Properties & Relations  (7)

Use VertexCount and EdgeCount to count vertices and edges:

Use VertexList and EdgeList to enumerate vertices and edges in standard order:

Compute the AdjacencyMatrix from a graph:

Distance metric generates undirected graphs:

Dissimilarity generates directed graphs:

Generate a nearest neighbor graph within radius 0.2 by using SpatialGraphDistribution:

Use Nearest and NearestFunction get the nearest neighbor graph:

This is equivalent to:

Introduced in 2015
 (10.2)