mirror of
https://github.com/Brandon-Rozek/website.git
synced 2024-11-13 20:27:29 -05:00
29 lines
No EOL
1.1 KiB
Markdown
29 lines
No EOL
1.1 KiB
Markdown
---
|
|
date: 2022-04-07 19:48:12-04:00
|
|
draft: false
|
|
math: false
|
|
medium_enabled: true
|
|
medium_post_id: e5c5330cc9a7
|
|
tags:
|
|
- Python
|
|
title: Networkx Random Sample Graph
|
|
---
|
|
|
|
I've been working on several algorithms in `networkx`. In order to speed up testing, especially on large graphs, I've been randomly sampling portions of the original graph. The best way I've found to do this is through the following python snippet:
|
|
|
|
```python
|
|
import random
|
|
random_sample_edges = random.sample(list(G.edges), SAMPLE_SIZE)
|
|
G_sample = nx.Graph()
|
|
G_sample.add_edges_from(random_sample_edges)
|
|
```
|
|
|
|
It might be tempting to sample the nodes and then grab the subgraph like the following:
|
|
|
|
```python
|
|
import random
|
|
random_nodes = random.sample(list(G.nodes), SAMPLE_SIZE)
|
|
G_sample = G.subgraph(random_nodes)
|
|
```
|
|
|
|
However, only considering the nodes when sampling makes it highly likely that the subgraph will significantly less edges. This results in a mostly disconnected subgraph and a loss of information. Sampling the edges prevents this issue at the expense of not capturing single nodes not connected to anything else. |