Basic usage

There are three common ways to obtain a graph: load a graph from file, generate a graph (igraph.generate for determinate generators, igraph.randgame for stochastic generators), or load a famous graph.

• igraph.load
• igraph.generate
• igraph.randgame
• igraph.famous

The primary function for modifying graphs is igraph.rewire. It is also possible to use builtin MATLAB functions for modifying a graph in the same manner as modifying matrices.

To characterize a graph there are a few functions including those that characterize nodes and those that predict community structure. The main function for characterizing nodes is igraph.centrality, which provides many methods for defining the centrality of a node. For network level analysis, igraph.cluster will group nodes into a membership vector. The resulting communities can be compared with igraph.compare or scored using igraph.modularity.

The igraph.plot function displays a graph in a MATLAB plot. This has options for viewing node membership by coloring the nodes and using size to show a node metric (such as one produced by igraph.centrality). Layouts can be generate directly via igraph.layout, if there is need to modify a layout or reuse a layout for multiple plots. Additionally, if representing a graph with MATLAB's builtin graph classes, the graph.plot() method can be used.

Once finished processing the graph, it may be useful to save it to file in order to use it outside of MATLAB or to prevent having to rerun the processing steps. Saving graphs can be done using the igraph.save function.

• MATLAB's builtin rng or igraph.rng can be used to ensure reproducible results when working with any indeterminate functions (See the RNG guide).
• igraph.isdirected and igraph.isweighted are used by many functions to guess characteristics of the input graph. If a function is not behaving as expected, it may be worthwhile to look at the help for these functions to understand how they determine whether a graph is directed or weighted. When in doubt, the type of graph can be explicitly stated in the function call.

To follow MATLAB convention, many of the functions in the igraph namespace provide access to multiple related igraph functions. This is similar to how a function like corr can be supplied a correlation method, Pearson, Spearman, or Kendall. For example, the igraph.load function provides access to all the igraph functions related to reading a graph from a file or the igraph.cluster function provides access to different community detection algorithms. These functions aim to select an intelligent default when applicable (i.e. igraph.load can guess the correct backend function based on the provided file extension) other functions may default to a simple method (igraph.compare defaults to normalized mutual information) or require a method to be provided (igraph.cluster). The defaults can be overridden by supplying a method argument. See the help for specific functions for which methods are available and any method specific options there are.

For igraph.load and igraph.save, file names are used to guess the file type based on the name's extension. The type can be explicitly provided using the FileType name-value pair. Available types are:

• mat
• edgelist
• dl
• ncol
• lgl
• graphdb
• graphml
• gml
• pajek
• dot
• leda

For more information about the file types see the igraph reference manual.

Functions that accept a matrix often need to know if the matrix is weighted or directed. The igraph.isweighted and igraph.isdirected functions are used to guess these. To override the guess use the isdirected or isweighted name-value pairs. Note: not all functions use weights, those the do not will not accept an isweighted value.

If a function accepts attributes (node or edge) these can be supplied either as a vector (with length equal to the number of nodes or edges) or as the name of an attribute attached to the graph. When using a matrix representation attributes must be supplied as a vector, for graph classes either a vector or a name can be used (see the attribute guide).

When a function returns a graph, it provides optional arguments for the representation of the resulting adjacency matrix. The repr option can be passed any of {"sparse", "full", "graph"} where "sparse" and "full" will return matrices and "graph" will return either a graph or digraph object depending on if the graph is directed or not. dtype can be used to determine the datatype, "logical" or "double".

For functions that accept and return a graph, the input graph will be used as a template, for example when using igraph.rewire if the input graph is a sparse matrix, the graph returned will also be a sparse matrix.