I read wiki (http://wiki.apache.org/nutch/NewScoring#LinkRank) stating the process of LinkRank is iterative and scores tend to converge after iteration. However, from the the source I discover it seems that the job always reads from the same input path and produce to the same output path. For instance,
runCounter() reads intput from nodes and returns the number of nodes runInitializer() reads from nodes and initializes inLinkScore then iteration (default is 10) runInverted() reads from nodes, where inLinkScore is initialized, outlinks, and loops; then produces output to linkrank-<random>/inverted runAnalysis() reads from nodes (inLinkScore is inited), and inverted path (in previous step); then produces output to linkrank-<random>/nodes This seems to me with the same process to calculate the scores, the result of LinkRank will always be the same at each iteration. So I can't understand very well how scores would converge. What place would be the key point to spot at? Or any doc that may explain this more detail? Thanks.
