From b2703a63380116d083c91a18a4018a393afff36f Mon Sep 17 00:00:00 2001
From: Renato Melo <renatinho.smelo@gmail.com>
Date: Mon, 21 Mar 2016 09:19:32 -0300
Subject: [PATCH] PrevalentSeed
---
src/algoritmos/MinDominatingSet.java | 2 +-
...DominatingSeed.java => PrevalentSeed.java} | 56 +++++++++++++++----
src/grafos/DirectedSocialNetwork.java | 46 ++++++++-------
3 files changed, 70 insertions(+), 34 deletions(-)
rename src/algoritmos/{DominatingSeed.java => PrevalentSeed.java} (82%)
diff --git a/src/algoritmos/MinDominatingSet.java b/src/algoritmos/MinDominatingSet.java
index b217840..c16e9a7 100644
--- a/src/algoritmos/MinDominatingSet.java
+++ b/src/algoritmos/MinDominatingSet.java
@@ -88,7 +88,7 @@ public class MinDominatingSet {
}
/**
- * Está errado, tem que olhar melhor a questão da direção das arestas
+ * Com grafo direcionado não retorna um conjunto dominante, mas é um bom cojunto de vértices influentes
**/
public HashSet<Actor> fastGreedy(DirectedSocialNetwork grafo) {
diff --git a/src/algoritmos/DominatingSeed.java b/src/algoritmos/PrevalentSeed.java
similarity index 82%
rename from src/algoritmos/DominatingSeed.java
rename to src/algoritmos/PrevalentSeed.java
index f62ac8b..e333ef9 100644
--- a/src/algoritmos/DominatingSeed.java
+++ b/src/algoritmos/PrevalentSeed.java
@@ -5,6 +5,10 @@ import java.util.PriorityQueue;
import java.util.Random;
import java.util.Set;
+import util.ComparaPorGrau;
+
+import com.google.common.collect.MinMaxPriorityQueue;
+
import algoritmos.MarginalGain;
import geradores.SocialNetworkGenerate;
import grafos.Actor;
@@ -13,33 +17,61 @@ import interfaces.SeedChooser;
/*De acordo com os experimentos, utilizando o CELF como subrotina
* dentro do conjunto dominante obtem-se um resultado semelhante
- * ao CELF original, o que leva a crer sãio fortes indicativos de
+ * ao CELF original, o que leva a crer são fortes indicativos de
* que os vértices "bons" estão dentro do conjunto dominante minimo
* num grafo direcionado
*/
-public class DominatingSeed implements SeedChooser<Actor> {
+public class PrevalentSeed implements SeedChooser<Actor> {
private DirectedSocialNetwork grafo;
- public DominatingSeed(DirectedSocialNetwork g) {
+ public PrevalentSeed(DirectedSocialNetwork g) {
this.grafo = g;
}
+
+ public HashSet<Actor> preSelecao(DirectedSocialNetwork grafo) {
+ int n = grafo.vertexSet().size();
+
+ // DS <-- {}
+ HashSet<Actor> candidatos = new HashSet<>();
+
+ // compare os vertices pelo grau
+ ComparaPorGrau comp = new ComparaPorGrau(grafo);
+
+ MinMaxPriorityQueue<Actor> heapMinMax = null;
+ heapMinMax = MinMaxPriorityQueue.orderedBy(comp).maximumSize(n)
+ .create();
+ for (Actor v : grafo.vertexSet()) {
+ heapMinMax.add(v);
+ }
+
+ Set<Actor> cobertos = new HashSet<>();
+ while (cobertos.size() < n) {
+
+ Actor v = heapMinMax.removeLast();
+ Set<Actor> vizinhos = grafo.outNeighborsOf(v);
+
+ if (cobertos.addAll(vizinhos)) {
+ candidatos.add(v);
+ }
+ cobertos.add(v);
+ }
+ return candidatos;
+ }
@Override
public HashSet<Actor> escolher(int k) {
HashSet<Actor> semente = new HashSet<Actor>();
- HashSet<Actor> minSet = new HashSet<Actor>();
+ HashSet<Actor> candidatos = new HashSet<Actor>();
- MinDominatingSet ds = new MinDominatingSet();
- minSet = ds.fastGreedy(grafo);
+ candidatos = preSelecao(grafo);
- System.out.println("|DS| = " + minSet.size());
- if (minSet.size() < k) {
- System.out.println("Erro: o cojunto domintante é menor que K");
+ if (candidatos.size() < k) {
+ System.out.println("Erro: o cojunto de candidatos é menor que K");
return null;
}
// create priority queue of all nodes, with marginal gain delta +inf
- PriorityQueue<MarginalGain> fila = priorityQueueOfGains(minSet);
+ PriorityQueue<MarginalGain> fila = priorityQueueOfGains(candidatos);
double MaxSpread = 0;
@@ -231,7 +263,7 @@ public class DominatingSeed implements SeedChooser<Actor> {
2.5);
long startTime = 0;
startTime = System.nanoTime();
- HashSet<Actor> seed = new DominatingSeed(g).escolher2(15);
+ HashSet<Actor> seed = new PrevalentSeed(g).escolher2(15);
System.out.println("Tempo: " + (System.nanoTime() - startTime) / 1000);
HashSet<Actor> ativos = g.indepCascadeDiffusion(seed);
@@ -243,7 +275,7 @@ public class DominatingSeed implements SeedChooser<Actor> {
// g.visualize();
startTime = System.nanoTime();
- HashSet<Actor> seed2 = new DominatingSeed(g).escolher(15);
+ HashSet<Actor> seed2 = new PrevalentSeed(g).escolher(15);
System.out.println("Tempo: " + (System.nanoTime() - startTime) / 1000);
HashSet<Actor> ativos2 = g.indepCascadeDiffusion(seed2);
diff --git a/src/grafos/DirectedSocialNetwork.java b/src/grafos/DirectedSocialNetwork.java
index f9d2f3a..fb86913 100644
--- a/src/grafos/DirectedSocialNetwork.java
+++ b/src/grafos/DirectedSocialNetwork.java
@@ -186,7 +186,7 @@ public class DirectedSocialNetwork extends
public double espectedSpread(HashSet<Actor> seed, boolean ic) {
double media = 0;
int soma = 0;
- int repeticoes = 10000;
+ int repeticoes = 5000;
if (!ic) {
HashSet<Actor> ativados = linearThresholdDiffusion(seed);
@@ -243,6 +243,7 @@ public class DirectedSocialNetwork extends
while (escolhidos.size() < k) {
Actor maior = heapMinMax.removeLast();
+ System.out.print(maior+"; ");
escolhidos.add(maior);
}
@@ -327,29 +328,32 @@ public class DirectedSocialNetwork extends
DirectedSocialNetwork g = new DirectedSocialNetwork(
DefaultWeightedEdge.class);
- g = new SocialNetworkGenerate().gerarGrafo(40, 2);
-
- HashSet<Actor> seed = new HashSet<>();
-
- Set<Actor> vertices = g.vertexSet();
- for (Actor actor : vertices) {
- if (seed.size() < 5) {
- seed.add(actor);
- }
-
- }
-
+ g = new SocialNetworkGenerate().gerarGrafo(30, 2);
+
+// HashSet<Actor> seed = new HashSet<>();
+//
+// Set<Actor> vertices = g.vertexSet();
+// for (Actor actor : vertices) {
+// if (seed.size() < 5) {
+// seed.add(actor);
+// }
+//
+// }
+//
System.out.println("|V(G)| = " + g.vertexSet().size());
System.out.println("|E(G)| = " + g.edgeSet().size());
- System.out.println("Seed:");
- for (Actor a : seed) {
- System.out.println(a.toString());
- }
-
- System.out.println("Difusão média no modelo IC = "+g.espectedSpread(seed, true));
- System.out.println("Difusão média no modelo LT = "+g.espectedSpread(seed, false));
+ System.out.println("Q: ");
+ Set<Actor> Q = g.verticesGrauMaior(g.vertexSet(), g.vertexSet().size());
+ System.out.println();
+// System.out.println("Seed:");
+// for (Actor a : seed) {
+// System.out.println(a.toString());
+// }
+//
+// System.out.println("Difusão média no modelo IC = "+g.espectedSpread(seed, true));
+// System.out.println("Difusão média no modelo LT = "+g.espectedSpread(seed, false));
// g.activate(seed);
-// g.visualize();
+ g.visualize();
}
}
--
GitLab