Last but not least, the PageRank algorithm introduces a damping factor to mitigate the effect of the neighbors. The idea is that, when navigating from one page to another, the internet user usually clicks on links. But in some cases, the user might become bored or, for whatever reason, go to another page. The probability of this happening is modeled by the damping factor. The final PageRank formula from the initial paper is as follows:
PR(A) = (1 - d) + d * (PR(N1)/C(N1) + ... + PR(Nn)/C(Nn))
Usually, the damping factor d is set to be around 0.85, meaning the probability of a user randomly navigating to a page without following a link is 1 - d = 15%.
Another important effect of the damping factor is visible for nodes with no outgoing links, also called sinks. Without the damping factor, these nodes will have the tendency to take the rank given by their neighbors without giving it back, breaking the principle of the algorithm.
