Refine the code according to coding convention.

shogun-toolbox · Apr 5, 2012 · 49b6e2e · 49b6e2e
1 parent 9473a5d
commit 49b6e2e
Show file tree

Hide file tree

Showing 5 changed files with 99 additions and 64 deletions.
diff --git a/src/shogun/evaluation/ClusteringAccuracy.cpp b/src/shogun/evaluation/ClusteringAccuracy.cpp
@@ -0,0 +1,26 @@
+/*
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 3 of the License, or
+ * (at your option) any later version.
+ *
+ * Written (W) 2012 Chiyuan Zhang
+ * Copyright (C) 2012 Chiyuan Zhang
+ */
+
+#include <shogun/evaluation/ClusteringAccuracy.h>
+
+using namespace shogun;
+
+float64_t CClusteringAccuracy::evaluate(CLabels* predicted, CLabels* ground_truth)
+{
+	SGVector<int32_t> predicted_ilabels=predicted->get_int_labels();
+	SGVector<int32_t> groundtruth_ilabels=ground_truth->get_int_labels();
+	int32_t correct=0;
+	for (int32_t i=0; i < predicted_ilabels.vlen; ++i)
+	{
+		if (predicted_ilabels[i] == groundtruth_ilabels[i])
+			correct++;
+	}
+	return float64_t(correct)/predicted_ilabels.vlen;
+}
diff --git a/src/shogun/evaluation/ClusteringAccuracy.h b/src/shogun/evaluation/ClusteringAccuracy.h
@@ -35,18 +35,7 @@ class CClusteringAccuracy: public CClusteringEvaluation
 	 * @param ground_truth labels assumed to be correct
 	 * @return evaluation result
 	 */
-	virtual float64_t evaluate(CLabels* predicted, CLabels* ground_truth)
-	{
-		SGVector<int32_t> predicted_ilabels=predicted->get_int_labels();
-		SGVector<int32_t> groundtruth_ilabels=ground_truth->get_int_labels();
-		int32_t correct=0;
-		for (int32_t i=0; i < predicted_ilabels.vlen; ++i)
-		{
-			if (predicted_ilabels[i] == groundtruth_ilabels[i])
-				correct++;
-		}
-		return float64_t(correct)/predicted_ilabels.vlen;
-	}
+	virtual float64_t evaluate(CLabels* predicted, CLabels* ground_truth);
 
 	/** @return whether criterium has to be maximized or minimized */
 	virtual EEvaluationDirection get_evaluation_direction()

diff --git a/src/shogun/evaluation/ClusteringEvaluation.h b/src/shogun/evaluation/ClusteringEvaluation.h
@@ -11,6 +11,8 @@
 #ifndef __CLUSTERINGEVALUATION_H__
 #define __CLUSTERINGEVALUATION_H__
 
+#include <vector>
+
 #include <shogun/evaluation/Evaluation.h>
 #include <shogun/features/Labels.h>
 

diff --git a/src/shogun/evaluation/ClusteringMutualInformation.cpp b/src/shogun/evaluation/ClusteringMutualInformation.cpp
@@ -0,0 +1,69 @@
+/*
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 3 of the License, or
+ * (at your option) any later version.
+ *
+ * Written (W) 2012 Chiyuan Zhang
+ * Copyright (C) 2012 Chiyuan Zhang
+ */
+
+#include <vector>
+#include <cmath>
+#include <algorithm>
+
+#include <shogun/evaluation/ClusteringMutualInformation.h>
+
+using namespace shogun;
+
+float64_t CClusteringMutualInformation::evaluate(CLabels* predicted, CLabels* ground_truth)
+{
+	std::vector<int32_t> label_p=unique_labels(predicted);
+	std::vector<int32_t> label_g=unique_labels(ground_truth);
+
+	if (label_p.size() != label_g.size())
+		SG_ERROR("Number of classes are different\n");
+	uint32_t n_class=label_p.size();
+	float64_t n_label=predicted->get_num_labels();
+
+	SGVector<int32_t> ilabels_p=predicted->get_int_labels();
+	SGVector<int32_t> ilabels_g=ground_truth->get_int_labels();
+
+	SGMatrix<float64_t> G(n_class, n_class);
+	for (size_t i=0; i < n_class; ++i)
+	{
+		for (size_t j=0; j < n_class; ++j)
+			G(i, j)=find_match_count(ilabels_g, label_g[i],
+				ilabels_p, label_p[j])/n_label;
+	}
+
+	std::vector<float64_t> G_rowsum(n_class), G_colsum(n_class);
+	for (size_t i=0; i < n_class; ++i)
+	{
+		for (size_t j=0; j < n_class; ++j)
+		{
+			G_rowsum[i] += G(i, j);
+			G_colsum[i] += G(j, i);
+		}
+	}
+
+	float64_t mutual_info = 0;
+	for (size_t i=0; i < n_class; ++i)
+	{
+		for (size_t j=0; j < n_class; ++j)
+		{
+			if (G(i, j) != 0)
+				mutual_info += G(i, j) * log(G(i,j) /
+					(G_rowsum[i]*G_colsum[j]))/log(2.);
+		}
+	}
+
+	float64_t entropy_p = 0, entropy_g = 0;
+	for (size_t i=0; i < n_class; ++i)
+	{
+		entropy_g += -G_rowsum[i] * log(G_rowsum[i])/log(2.);
+		entropy_p += -G_colsum[i] * log(G_colsum[i])/log(2.);
+	}
+
+	return mutual_info / std::max(entropy_g, entropy_p);
+}
diff --git a/src/shogun/evaluation/ClusteringMutualInformation.h b/src/shogun/evaluation/ClusteringMutualInformation.h
@@ -11,10 +11,6 @@
 #ifndef __CLUSTERINGMUTUALINFORMATION_H__
 #define __CLUSTERINGMUTUALINFORMATION_H__
 
-#include <vector>
-#include <cmath>
-#include <algorithm>
-
 #include <shogun/evaluation/ClusteringEvaluation.h>
 
 namespace shogun
@@ -39,54 +35,7 @@ class CClusteringMutualInformation: public CClusteringEvaluation
 	 * @param ground_truth labels assumed to be correct
 	 * @return evaluation result
 	 */
-	virtual float64_t evaluate(CLabels* predicted, CLabels* ground_truth)
-	{
-		std::vector<int32_t> label_p=unique_labels(predicted);
-		std::vector<int32_t> label_g=unique_labels(ground_truth);
-
-		if (label_p.size() != label_g.size())
-			SG_ERROR("Number of classes are different\n");
-		uint32_t n_class=label_p.size();
-		float64_t n_label=predicted->get_num_labels();
-
-		SGVector<int32_t> ilabels_p=predicted->get_int_labels();
-		SGVector<int32_t> ilabels_g=ground_truth->get_int_labels();
-
-		SGMatrix<float64_t> G(n_class, n_class);
-		for (size_t i=0; i < n_class; ++i)
-			for (size_t j=0; j < n_class; ++j)
-				G(i, j)=find_match_count(ilabels_g, label_g[i],
-					ilabels_p, label_p[j])/n_label;
-		std::vector<float64_t> G_rowsum(n_class), G_colsum(n_class);
-		for (size_t i=0; i < n_class; ++i)
-		{
-			for (size_t j=0; j < n_class; ++j)
-			{
-				G_rowsum[i] += G(i, j);
-				G_colsum[i] += G(j, i);
-			}
-		}
-
-		float64_t mutual_info = 0;
-		for (size_t i=0; i < n_class; ++i)
-		{
-			for (size_t j=0; j < n_class; ++j)
-			{
-				if (G(i, j) != 0)
-					mutual_info += G(i, j) * log(G(i,j) / 
-						(G_rowsum[i]*G_colsum[j]))/log(2.);
-			}
-		}
-
-		float64_t entropy_p = 0, entropy_g = 0;
-		for (size_t i=0; i < n_class; ++i)
-		{
-			entropy_g += -G_rowsum[i] * log(G_rowsum[i])/log(2.);
-			entropy_p += -G_colsum[i] * log(G_colsum[i])/log(2.);
-		}
-
-		return mutual_info / std::max(entropy_g, entropy_p);
-	}
+	virtual float64_t evaluate(CLabels* predicted, CLabels* ground_truth);
 
 	/** @return whether criterium has to be maximized or minimized */
 	virtual EEvaluationDirection get_evaluation_direction()