08.10.2016 Views

Foundations of Data Science

2dLYwbK

2dLYwbK

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Exercise 4.61 Consider an n × n grid in the plane.<br />

1. Prove that for any vertex u, there are at least k vertices at distance k for 1 ≤ k ≤<br />

n/2.<br />

2. Prove that for any vertex u, there are at most 4k vertices at distance k.<br />

3. Prove that for one half <strong>of</strong> the pairs <strong>of</strong> points, the distance between them is at least<br />

n/4.<br />

Exercise 4.62 Recall the definition <strong>of</strong> a small-world graph in Section 4.10. Show that<br />

in a small-world graph with r ≤ 2, that there exist short paths with high probability. The<br />

pro<strong>of</strong> for r = 0 is in the text.<br />

Exercise 4.63 Change the small worlds graph as follows. Start with a n × n grid where<br />

each vertex has one long-distance edge to a vertex chosen uniformly at random. These are<br />

exactly like the long-distance edges for r = 0. But the grid edges are not present. Instead,<br />

we have some other graph with the property that for each vertex, there are Θ(t 2 ) vertices<br />

at distance t from the vertex for t ≤ n. Show that, almost surely, the diameter is O(ln n).<br />

Exercise 4.64 Consider an n-node directed graph with two random out-edges from each<br />

node. For two vertices s and t chosen at random, prove that with high probability there<br />

exists a path <strong>of</strong> length at most O(ln n) from s to t.<br />

Exercise 4.65 Explore the concept <strong>of</strong> small world by experimentally determining the answers<br />

to the following questions:<br />

1. How many edges are needed to disconnect a small world graph? By disconnect we<br />

mean at least two pieces each <strong>of</strong> reasonable size. Is this connected to the emergence<br />

<strong>of</strong> a giant component?<br />

2. How does the diameter <strong>of</strong> a graph consisting <strong>of</strong> a cycle change as one adds a few<br />

random long distance edges?<br />

Exercise 4.66 In the small world model, would it help if the algorithm could look at edges<br />

at any node at a cost <strong>of</strong> one for each node looked at?<br />

Exercise 4.67 Consider the n × n grid in the section on small world graphs. If the<br />

probability <strong>of</strong> an edge from vertex u to vertex v is proportional to d −r (u, v), show that the<br />

constant <strong>of</strong> proportionality c r (u) is<br />

θ(n 2−r ) for r > 2<br />

θ(ln n) for r = 2<br />

θ(1) for r < 2<br />

Exercise 4.68 Show that for r < 2 in the small world graph model that short paths exist<br />

but a polylog length path is unlikely to encounter a long distance edge whose end point is<br />

close to the destination.<br />

Exercise 4.69 Make a list <strong>of</strong> the ten most interesting things you learned about random<br />

graphs.<br />

138

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!