README.Rmd

---
title: "Make Reduced From Partition"
author: "Steven Wolf"
date: "`r Sys.Date()`"
output: github_document
---

## Setup for the new function `make_reduced_from_partition`
Load all libraries and load the new function 

```{r}
library(concorR)
library(igraph)
library(viridis)

#source('mrfp.R')
```
This function ~~will be~~ has been added to the `concorR` repository on my fork ~~in the next few days~~.  The Travis status is:

<!-- badges: start -->
[![Travis build status](https://travis-ci.com/sfwolfphys/concorR.svg?branch=main)](https://travis-ci.com/sfwolfphys/concorR)
<!-- badges: end -->

## Proof of concept
Note:  given the fact that we have an existing network/partition, I decided to make the input be a single adjacency matrix/partition rather than a list of adjacency matricies/partitions.  We want to have the list input for the other method because concor uses list inputs.  However, it is superfluous here.

I messed with `edge_betweenness` to find communities in the examples in the standard README for `concorR`.  They were not instructive (only one community), so I abandoned that.  Instead, I have created a random network and found communities there:
```{r}
set.seed(1234)
g  <-  erdos.renyi.game(50,p=0.2)
g_adj  <- as.matrix(as_adjacency_matrix(g))
ebc.g  <- edge.betweenness.community(g)
ebPart  <- ebc.g$membership
```

Show the whole network:
```{r}
vColor = viridis(max(unlist(ebPart)))
plot(g, vertex.color = vColor[ebc.g$membership])
```


First, let's show that the reduced network works for the degree statistic:
```{r}
g.red  <- make_reduced_from_partition(g_adj, ebPart, stat='degree')
plot_reduced(make_reduced_igraph(g.red$reduced_mat))
```

And now, the density statistic:
```{r}
g.red.den  <- make_reduced_from_partition(g_adj, ebPart, stat='density')
plot_reduced(make_reduced_igraph(g.red.den$reduced_mat))
```

## Full function text
```{r}
make_reduced_from_partition
```