SNAP複雜網路處理平臺

簡介：這是斯坦福提供的，用於處理複雜網路的平臺庫，核心是使用C++編寫的，效率相對很高，提供了C++和Python的介面，原始碼可以在snap網站下載，也可以在github上下載。

C++

編譯，安裝與測試

需要安裝兩個相關的軟體，用於繪圖，Gnuplot和Graphviz這兩個的安裝都很簡單，在ubuntu上只需要一行安裝命令即可，具體的可百度。

在ubuntu環境下，在snap根目錄下執行make all，執行完成之後，進入到example資料夾下，可以使用已經提供的介面，例如，使用下列命令生成一個小世界網路。

cd examples/graphgen 
./graphgen -g:w -n:1000 -k:4 -p:0.1 -o:smallworld.txt
 

生成指定度分佈的冪律圖

檢視檔案，發現只需要提供nodes和exponent即可生成，速度極快，可能採用的是文章《Efficient and Simple Generation of Random Simple Connected Graphs with Prescribed Degree Sequence》類似的方法，使用命令為：

./graphgen -g:p -n:100000 -p:1.5 -o:t1.5.txt

-n表示nodes數

-p表示exponent（指數）

-o表示輸出檔案

檢視原始碼，在examples中的graphgen.cpp中，可以使用已經提供的生成圖方式。生成冪律圖對應以下函式。

G = TSnap::GenRndPowerLaw(N, P, true);

經測試，生成的圖符合分佈！

附graphgen.cpp檔案原始碼

#include "stdafx.h"

int main(int argc, char* argv[]) {
  Env = TEnv(argc, argv, TNotify::StdNotify);
  Env.PrepArgs(TStr::Fmt("Graph generators. build: %s, %s. Time: %s", __TIME__, __DATE__, TExeTm::GetCurTm()));
  TExeTm ExeTm;
 

  Try
  const TStr OutFNm = Env.GetIfArgPrefixStr("-o:", "output.txt", "Output graph filename");
  const TStr Plot = Env.GetIfArgPrefixStr("-g:", "e", "Which generator to use:"
    "\n\tf: Complete graph. Required parameters: n (number of nodes)"
    "\n\ts: Star graph. Required parameters: n (number of nodes)"
    "\n\t2: 2D Grid. Required parameters: n (number of rows), m (number of columns)"
    "\n\te: Erdos-Renyi (G_nm). Required parameters: n (number of nodes), m (number of edges)"
    "\n\tk: Random k-regular graph. Required parameters: n (number of nodes), k (degree of every node)"
    "\n\tb: Albert-Barabasi Preferential Attachment. Required parameters: n (number of nodes), k (edges created by each new node)"
    "\n\tp: Random Power-Law graph. Required parameters: n (number of nodes), p (power-law degree exponent)"
    "\n\tc: Copying model by Kleinberg et al. Required parameters: n (number of nodes), p (copying probability Beta)"
    "\n\tw: Small-world model. Required parameters: n (number of nodes), k (each node is connected to k nearest neighbors in ring topology), p (rewiring probability)\n"
    );
  const int N = Env.GetIfArgPrefixInt("-n:", 1000, "Number of nodes");
  const int M = Env.GetIfArgPrefixInt("-m:", 5000, "Number of edges");
  const double P = Env.GetIfArgPrefixFlt("-p:", 0.1, "Probability/Degree-exponent");
  const int K = Env.GetIfArgPrefixInt("-k:", 3, "Degree");

  if (Env.IsEndOfRun()) { return 0; }
  TExeTm ExeTm;
  TInt::Rnd.PutSeed(0); // initialize random seed
  printf("Generating...\n");
  PUNGraph G;
  TStr DescStr;
  if (Plot == "f") {
    G = TSnap::GenFull<PUNGraph>(N);
    DescStr = TStr::Fmt("Undirected complete graph.");
  } else
  if (Plot == "s") {
    G = TSnap::GenStar<PUNGraph>(N, false);
    DescStr = TStr::Fmt("Undirected star graph (1 center node connected to all other nodes).");
  } else
  if (Plot == "2") {
    G = TSnap::GenGrid<PUNGraph>(N, M, false);
    DescStr = TStr::Fmt("Undirected 2D grid of %d rows and %d columns.", N, M);
  } else
  if (Plot == "e") {
    G = TSnap::GenRndGnm<PUNGraph>(N, M, false);
    DescStr = TStr::Fmt("Undirected Erdos-Renyi random graph.");
  } else
  if (Plot == "k") {
    G = TSnap::GenRndDegK(N, K);
    DescStr = TStr::Fmt("Undirected k-regular random graph (every node has degree K).");
  } else
  if (Plot == "b") {
    G = TSnap::GenPrefAttach(N, K);
    DescStr = TStr::Fmt("Undirected Albert-Barabasi Preferential Attachment graph (each new node creades k preferentially attached edges).");
  } else
  if (Plot == "p") {
    G = TSnap::GenRndPowerLaw(N, P, true);
    DescStr = TStr::Fmt("Random Graph with Power-Law degree distribution with exponent P.");
  } else
  if (Plot == "c") {
    G = TSnap::ConvertGraph<PUNGraph>(TSnap::GenCopyModel(N, P));
    DescStr = TStr::Fmt("Copying model by Kleinberg et al. Node u comes, selects a random v, and with prob P it links to v, with 1-P links u links to neighbor of v. Power-law degree slope is 1/(1-P).");
  } else
  if (Plot == "w") {
    G = TSnap::GenSmallWorld(N, K, P);
    DescStr = TStr::Fmt("Watts-Strogatz Small-world model. Every node links to K other nodes.");
  }
  printf("done.\n");
  TSnap::SaveEdgeList(G, OutFNm, DescStr);

  Catch
  printf("\nrun time: %s (%s)\n", ExeTm.GetTmStr(), TSecTm::GetCurTm().GetTmStr().CStr());
  return 0;
}

Python

與C++僅介面的呼叫形式不同，核心實現使用的是C++

重點總結

SNAP這個平臺一直以來沒有發現，這應該是非常優秀的一個平臺，速度極快，有時間可以好好研究一下。