5

我正在尝试实现一个非常简单的优先附件算法来创建无标度网络。它们具有遵循幂律的度数分布,即 P(k) ~ k^-g,其中 g 是指数。下面的算法应该产生指数等于 3 +/- 0.1 的度数分布,我的实现没有指数接近 2.5 +/- 0.1。我显然不理解某处的东西并继续出错。

很抱歉,如果这是在错误的地方,我无法决定它是否应该在 stackoverflow 或 maths.stackexchange.com 中。

The Algorithm:
Input: Number of Nodes N; Minimum degree d >= 1.
Output: scale-free multigraph
G = ({0,....,N-1}, E)
M: array of length 2Nd
for (v=0,...,n-1)
   for (i=0,...,d-1)
      M[2(vd+i)] = v;
      r = random number selected uniformly at random from {0,.....,2(vd+i)};
      M[2(vd+i)+1] = M[r];
   end
end

E = {};
for (i=0,...,nd-1)
   E[i] = {M[2i], M[2i+1]}
end

我在 C/C++ 中的实现:

void SF_LCD(std::vector< std::vector<int> >& graph, int N, int d) {
    if(d < 1 || d > N - 1) {
        std::cerr << "Error: SF_LCD: k_min is out of bounds: " << d;
    }

    std::vector<int> M;
    M.resize(2 * N * d);

    int r = -1;
    //Use Batagelj's implementation of the LCD model
    for(int v = 0; v < N; v++) {
        for(int i = 0; i < d; i++) {
            M[2 * (v * d + i)] = v;
             r = mtr.randInt(2 * (v * d + i));
            M[2 * (v * d + i) + 1] = M[r];
        }
    }

    //create the adjacency list
    graph.resize(N);
    bool exists = false;
    for(int v = 0; v < M.size(); v += 2) {
        int m = M[v];
        int n = M[v + 1];

        graph[m].push_back(n);
        graph[n].push_back(m);
    }
}

这是我在 N = 10,000 和 d = 1 时得到的度数分布示例:

1   6674
2   1657
3   623
4   350
5   199
6   131
7   79
8   53
9   57
10  27
11  17
12  20
13  15
14  12
15  5
16  8
17  5
18  10
19  7
20  6
21  5
22  6
23  4
25  4
26  2
27  1
28  6
30  2
31  1
33  1
36  2
37  2
43  1
47  1
56  1
60  1
63  1
64  1
67  1
70  1
273 1
4

1 回答 1

6

好的,所以我不知道如何使这个特定算法正常工作,而是使用了另一个算法。

The Algorithm:
Input: Number of Nodes N; 
       Initial number of nodes m0; 
       Offset Exponent a; 
       Minimum degree 1 <= d <= m0.
Output: scale-free multigraph G = ({0,....,N-1}, E).

1) Add m0 nodes to G.
2) Connect every node in G to every other node in G, i.e. create a complete graph.
3) Create a new node i.
4) Pick a node j uniformly at random from the graph G. Set P = (k(j)/k_tot)^a.
5) Pick a real number R uniformly at random between 0 and 1.
6) If P > R then add j to i's adjacency list.
7) Repeat steps 4 - 6 until i has m nodes in its adjacency list.
8) Add i to the adjacency list of each node in its adjacency list.
9) Add i to to the graph.
10) Repeat steps 3 - 9 until there are N nodes in the graph.

其中 k(j) 是图 G 中节点 j 的度数,k_tot 是图 G 中边数(总度数)的两倍。

通过改变参数可以控制度数分布的指数。a = 1.22 给我一个 3 +/- 0.1 的指数 g(在 P(k) ~ k^-g 中)。

于 2012-05-17T12:06:34.017 回答