Implementing PageRank using MapReduce

Question

I\'m trying to get my head around an issue with the theory of implementing the PageRank with MapReduce.

I have the following simple scenario with three nodes: A B C.

Answer 1

A detailed explanation with Python code, by Michael Nielsen.

Answer 2

Here is a pseudocode:

map( key: [url, pagerank], value: outlink_list )
    for each outlink in outlink_list
        emit( key: outlink, value: pagerank/size(outlink_list) )

    emit( key: url, value: outlink_list )

reducer( key: url, value: list_pr_or_urls )
    outlink_list = []
    pagerank = 0

    for each pr_or_urls in list_pr_or_urls
        if is_list( pr_or_urls )
            outlink_list = pr_or_urls
        else
            pagerank += pr_or_urls

    pagerank = 1 - DAMPING_FACTOR + ( DAMPING_FACTOR * pagerank )

    emit( key: [url, pagerank], value: outlink_list )

It is important that in the reduce you should output outlinks and not inlinks, as some articles on the intenret suggests. This way the consecutive iterations will also have outlinks as input of the mapper.

Pay attention that multiple outlinks with the same address from the same page count as one. Also, don't count loops (page linking to itself).

The damping factor is traditionally 0.85, although you can play around with other values, too.

Answer 3

We iteratively evaluate PR. PR(x) = Sum(PR(a)*weight(a), a in in_links) by

map ((url,PR), out_links) //PR = random at start
for link in out_links
   emit(link, ((PR/size(out_links)), url))

reduce(url, List[(weight, url)):
   PR =0
   for v in weights
       PR = PR + v
   Set urls = all urls from list

   emit((url, PR), urls)

so the output equals input and we can do this until coverage.