我的网页排名之和为0.9收敛

问题描述

当我计算一组爬域名的网页排名，使用0.85阻尼因子。正如在许多页提到的行列论文，页面的PR值的总和应收敛于1。但不管有多少反复做，这似乎收敛于0.90xxx。如果我低衰减系数为0.5，我靠拢1显而易见的。

When I'm calculating page ranks of a set of crawled domains, using a dampening factor of 0.85. As mentioned in many page ranks papers, the sum of pageranks should converge to 1. But regardless of how many iterations I do, it seems to converge at 0.90xxx. If I lower dampening factor to 0.5, I move closer to 1 obviously.

是不是坏了网页排名总和收敛于0.90，什么会，这通常牵连？

Is it bad that the page ranks sum converge at 0.90, and what would this generally implicate?

推荐答案

这成为了算法：

// data structures

private HashMap<String, Double> pageRanks;
private HashMap<String, Double> oldRanks;
private HashMap<String, Integer> numberOutlinks;
private HashMap<String, HashMap<String, Integer>> inlinks;
private HashSet<String> domainsWithNoOutlinks;
private double N;

// data parsing occluded

public void startAlgorithm() {
    int maxIterations = 20;
    int itr = 0;
    double d = 0.85;
    double dp = 0;
    double dpp = (1 - d) / N;

    // initialize pagerank
    for (String s : oldRanks.keySet()) {
        oldRanks.put(s, 1.0 / N);
    }

    System.out.println("Starting page rank iterations..");
    while (maxIterations >= itr) {
        System.out.println("Iteration: " + itr);
        dp = 0;
        // teleport probability
        for (String domain : domainsWithNoOutlinks) {
            dp = dp + d * oldRanks.get(domain) / N;
        }
        for (String domain : oldRanks.keySet()) {
            pageRanks.put(domain, dp + dpp);

            for (String inlink : inlinks.get(domain).keySet()) { // for every inlink of domain
                pageRanks.put(domain, pageRanks.get(domain) + inlinks.get(domain).get(inlink) * d * oldRanks.get(inlink) / numberOutlinks.get(inlink));
            }
        }
        // update pageranks with new values
        for (String domain : pageRanks.keySet()) {
            oldRanks.put(domain, pageRanks.get(domain));
        }

        itr++;
    }
}

这条线是重要的：

Where this line is the important one:

pageRanks.put(domain, pageRanks.get(domain) + inlinks.get(domain).get(inlink) * d * oldRanks.get(inlink) / numberOutlinks.get(inlink));

inlinks.get（域）获得（内链接）返回多少的内链接如/引用当前域，我们除以多少反向链接，目前域名拥有。而inlinks.get（域）获得（内链接）是我错过了我的算法因此为什么总和不收敛于1。

inlinks.get(domain).get(inlink) returns how much an inlink "like/referenced" the current domain, and we divide that by how many inlinks that current domain have. And "inlinks.get(domain).get(inlink)" is what I missed in my algorithm hence why the sum didn't converge at 1.

了解更多： http://www.ccs.northeastern.edu /home/daikeshi/notes/PageRank.pdf

这篇关于我的网页排名之和为0.9收敛的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持！