I'm working with a graph in networkx, and I want to do pagerank with special emphasis on emphasizing some nodes over others. Reading over the documentation there seems to be a few parameters to enable this. One thing is nstart, where I can pass initial weightings to the algorithm. The other is personalization, so that the random walk is more likely to stay on some nodes than others.
What are the tradeoffs in terms of using one over the other?
2.1m questions
2.1m answers
60 comments
57.0k users