Add ability to generate graphs based on correlations of sequences

[rkingan]

Given a list of labeled real-valued vectors and a significance level alpha, generate a graph with vertices corresponding to the list's labels where two vertices are connected by an edge if the corresponding vectors have a correlation coefficient that is positive or negative at the specified significance level. An object creating the graph should support the plot method, which will plot the generated graph using a circular layout.

CC: @dcoudert

Component: graph theory

Keywords: sagedays@icerm

Author: Robert Kingan

Issue created by migration from https://trac.sagemath.org/ticket/25933

[rkingan]

comment:1

Added branch name

[rkingan]

Branch: correlation-graph

[Bruno-TT]

Commit: abc6100

[Bruno-TT]

comment:4

I've made a very bare-bones start on this but please tear it apart.

One note about my plot() implementation - the graph.plot function invokes the constructor at .__class__() about 10 methods deep (specifically, at the start of GenericGraph._subgraph_by_adding).

In the plot method I've put in a hack that works around it,

def plot(self):

    self.__class__=Graph
    p=self.plot()
    self.__class__=CorrelationGraph

    return p

but I think this is a symptom of a larger issue. Making assumptions about the arguments of the class constructor of a passed object feels like poor design, and we should be able to inherit the plot method with any class, without restrictions on the shape of said class' constructor.

If anyone can find a better workaround please let me know, this is the tidiest one I've found.

---

New commits:

abc6100

initial implementation start

[Bruno-TT]

Changed branch from correlation-graph to u/gh-Bruno-TT/correlation-graph

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

2798d32

base test fix

[sagetrac-git]

Changed commit from abc6100 to 2798d32

[dcoudert]

comment:7

Several comments:

• graph generators are usually placed in src/sage/graphs/generators/ and made accessible as graphs.<TAB>
• Your method must be commented. I'm not sure to understand the steps.
• If you think the current design of the plot method must be improved, please open a dedicated ticket

[sagetrac-git]

Changed commit from 2798d32 to ee12eac

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

ee12eac

changed class to generator method + tab completion

[dcoudert]

comment:9

Better this way.

Note that the current doctest is not working since parameter include_anticorrelation is required. You should certainly set its default value and add doctests using it.

Coding style:

• boolean_adjacency_matrix = (corrs>=alpha) -> boolean_adjacency_matrix = corrs >= alpha. I don't think you need the parathenses.
• adjacency_matrix=Matrix(boolean_adjacency_matrix.astype(int)) -> adjacency_matrix = Matrix(boolean_adjacency_matrix.astype(int))

and of course add the documentation.

[Bruno-TT]

PR here: #35009

I've changed the above as per your suggestions David, and added some documentation

[mkoeppe]

Removed branch from issue description; replaced by PR #35009