I've been recently working with graph sampling, and I can't seem to find useful explanation of the following two aspects. On one side there are pagerank-based algorithms, which converge to a stationary distribution, and are extremely useful for modern graph mining tasks.
On the other hand, many network embedding methods leverage $1^{st}$ or $2^{nd}$ order random walks for network sampling.
I am wondering what is the exact connection between the two? Is pagerank actually a form of a $2^{nd}$ order graph walk?
Thank you!