blog/ib002/10-graphs/bfs-tree.md

150 lines
3.1 KiB
Markdown
Raw Normal View History

---
title: Distance boundaries from BFS tree on undirected graphs
description: |
Short explanation of distance boundaries deduced from a BFS tree.
---
## Introduction
As we have talked on the seminar, if we construct from some vertex $u$ BFS tree on an undirected graph, we can obtain:
- lower bound of length of the shortest path between 2 vertices from the _height difference_
- upper bound of length of the shortest path between 2 vertices from the _path through the root_
## Lower bound
Consider the following graph:
![BFS graph](/files/ib002/bfs-tree/bfs_graph.svg)
<details>
<summary>Source for the rendered graph</summary>
```dot showLineNumbers
graph {
a -- c
a -- e
c -- i
c -- b
e -- j
i -- d
b -- h
d -- h
h -- j
}
```
</details>
We run BFS from the vertex $a$ and obtain the following BFS tree:
![BFS tree](/files/ib002/bfs-tree/bfs_tree.svg)
<details>
<summary>Source for the BFS tree</summary>
```dot showLineNumbers
digraph {
a -> c
a -> e
c -> b
c -> i
e -> j
b -> h
i -> d
}
```
</details>
Let's consider pair of vertices $e$ and $h$. For them we can safely lay, from the BFS tree, following properties:
- lower bound: $2$
- upper bound: $4$
By having a look at the graph we started from, we can see that we have a path $e, j, h$ that has a length 2. Apart from that we can also notice there is another path from $e$ to $h$ and that is $e, a, c, i, d, h$. And that path has a length of $5$. Doesn't this break our statements at the beginning? (_I'm leaving that as an exercise ;)_)
## Proof by contradiction
Let's keep the same graph, but break the lower bound, i.e. I have gotten a lower bound $2$, but „there must be a shorter path“! ;)
Now the more important question, is there a shorter path in that graph? The answer is no, there's no shorter path than the one with length $2$. So what can we do about it? We'll add an edge to have a shorter path. Now we have gotten a lower bound of $2$, which means the only shorter path we can construct has $1$ edge and that is $e, h$ (no intermediary vertices). Let's do this!
![BFS tree](/files/ib002/bfs-tree/bfs_graph_with_additional_edge.svg)
<details>
<summary>Source for the BFS graph with an additional edge</summary>
```dot showLineNumbers
graph {
a -- c
a -- e
c -- i
c -- b
e -- j
e -- h
i -- d
b -- h
d -- h
h -- j
}
```
</details>
Okay, so we have a graph that breaks the rule we have laid. However, we need to run BFS to obtain the new BFS tree, since we have changed the graph.
:::tip
Do we need to run BFS after **every** change?
­I am leaving that as an exercise ;)
:::
![BFS tree](/files/ib002/bfs-tree/bfs_tree_with_additional_edge.svg)
<details>
<summary>Source for the BFS tree</summary>
```dot showLineNumbers
digraph {
a -> c
a -> e
c -> b
c -> i
e -> h
e -> j
i -> d
}
```
</details>
Oops, we have gotten a new BFS tree, that has a height difference of 1.
:::tip
Try to think about a way this can be generalized for shortening of minimal length 3 to minimal length 2 ;)
:::