[mlpack-svn] r10252 - mlpack/trunk/src/mlpack/methods/pca

[fastlab-svn at coffeetalk-1.cc.gatech.edu]

Author: ajinkya
Date: 2011-11-11 16:29:55 -0500 (Fri, 11 Nov 2011)
New Revision: 10252

Modified:
   mlpack/trunk/src/mlpack/methods/pca/pca.cpp
   mlpack/trunk/src/mlpack/methods/pca/pca.hpp
   mlpack/trunk/src/mlpack/methods/pca/pca_test.cpp
Log:
added comments and made some formatting changes

Modified: mlpack/trunk/src/mlpack/methods/pca/pca.cpp
===================================================================
--- mlpack/trunk/src/mlpack/methods/pca/pca.cpp	2011-11-11 21:11:45 UTC (rev 10251)
+++ mlpack/trunk/src/mlpack/methods/pca/pca.cpp	2011-11-11 21:29:55 UTC (rev 10252)
@@ -1,3 +1,9 @@
+/**
+ * @file pca.cpp
+ *
+ * Implementation of PCA class to perform Principal Components Analysis on the
+ * specified data set.
+ */
 #include "pca.hpp"
 #include <mlpack/core.h>
 
@@ -4,16 +10,22 @@
 namespace mlpack {
 namespace pca {
 
-	PCA::PCA(){}
+PCA::PCA()
+{
+}
 
-	void PCA::Apply(arma::mat& coeff, arma::mat& score, arma::mat& data)
-	{
-		arma::princomp(coeff, score, arma::trans(data));
-		/*arma::vec eigval;
-		arma::mat eigvec;
-		arma::mat cov_mat = arma::cov(m_data);
-		arma::eig_sym(eigval, eigvec, cov_mat);
-		data_transformed = arma::trans(eigvec) * m_data;*/
-	}
-};
-};
+/**
+ * Apply Armadillo's Principal Component Analysis on the data set.
+ */
+void PCA::Apply(const arma::mat& data, arma::mat& coeff, arma::mat& score)
+{
+  //Armadillo's PCA api
+  arma::princomp(coeff, score, arma::trans(data));
+}
+
+PCA::~PCA()
+{
+}
+
+}; // namespace mlpack
+}; // namespace pca

Modified: mlpack/trunk/src/mlpack/methods/pca/pca.hpp
===================================================================
--- mlpack/trunk/src/mlpack/methods/pca/pca.hpp	2011-11-11 21:11:45 UTC (rev 10251)
+++ mlpack/trunk/src/mlpack/methods/pca/pca.hpp	2011-11-11 21:29:55 UTC (rev 10252)
@@ -1,3 +1,9 @@
+/**
+ * @file pca.hpp
+ *
+ * Defines the PCA class to perform Principal Components Analysis on the
+ * specified data set.
+ */
 #include <mlpack/core.h>
 namespace mlpack {
 namespace pca {
@@ -2,10 +8,24 @@
 
-class PCA {
+class PCA
+{
+ public:
+  PCA();
 
-	public:
-		PCA();
-		void Apply(arma::mat& coeff, arma::mat& score, arma::mat& data);
+  /**
+   * Apply Armadillo's Principal Component Analysis to the provided data set.
+   *
+   * @param data - Data matrix
+   * @param coeff - PCA Loadings
+   * @param score - contains the coordinates of the original data in the new coordinate system defined by the principal components
+   */
+  void Apply(const arma::mat& data, arma::mat& coeff, arma::mat& score);
 
-};
-};
-};
+  /**
+   * Delete PCA object
+   */
+  ~PCA();
+
+}; // class PCA
+
+}; // namespace pca
+}; // namespace mlpack

Modified: mlpack/trunk/src/mlpack/methods/pca/pca_test.cpp
===================================================================
--- mlpack/trunk/src/mlpack/methods/pca/pca_test.cpp	2011-11-11 21:11:45 UTC (rev 10251)
+++ mlpack/trunk/src/mlpack/methods/pca/pca_test.cpp	2011-11-11 21:29:55 UTC (rev 10252)
@@ -1,3 +1,9 @@
+/**
+ * @file pca_test.cpp
+ *
+ * Test file for PCA class
+ *
+ */
 #include "pca.hpp"
 
 using namespace std;
@@ -7,51 +13,49 @@
 
 int main(int argc, char *argv[])
 {
-	int n_rows;
-	int n_cols;
+  int n_rows;
+  int n_cols;
 
-	mat data;
-	data	<< 0.8402 << 0.7984 << 0.3352 << endr
-			<< 0.3944 << 0.9116 << 0.7682 << endr
-			<< 0.7831 << 0.1976 << 0.2778 << endr;
+  //input data set
+  mat data;
+  data << 0.8402 << 0.7984 << 0.3352 << endr
+      << 0.3944 << 0.9116 << 0.7682 << endr
+      << 0.7831 << 0.1976 << 0.2778 << endr;
 
-	mat coeff_actual;
+  //PCA loadings/coefficient matrix
+  mat coeff_actual;
+  coeff_actual << -0.41721 << -0.610904 <<  0.672854 << endr
+        <<  0.743785 << -0.654959 << -0.133446 << endr
+	<<  0.522226 <<  0.444775 <<  0.727636 << endr;
 
-	coeff_actual	<< -0.41721 << -0.610904 <<  0.672854 << endr
-					<<  0.743785 << -0.654959 << -0.133446 << endr
-					<<  0.522226 <<  0.444775 <<  0.727636 << endr;
+  //PCA score matrix for transformed
+  mat score_actual;
+  score_actual << -0.0144 << -0.2645 << 0 << endr
+        <<  0.4819 <<  0.1262 << 0 << endr
+        << -0.4675 <<  0.1383 << 0 << endr;
 
-	mat score_actual;
+  mat coeff;
+  mat score;
 
-	score_actual	<< -0.0144 << -0.2645 << 0 << endr
-					<<  0.4819 <<  0.1262 << 0 << endr
-					<< -0.4675 <<  0.1383 << 0 << endr;
+  //MLPACK data set is organized with rows as attributes and columns as data
+  //points. Hence the need to transpose the matrix before applying Armadillo's
+  //princomp PCA api
+  mat trans_data = arma::trans(data);
+  mlpack::pca::PCA p;
+  p.Apply(trans_data, coeff, score);
 
-	mat coeff;
-	mat score;
+  n_rows = data.n_rows;
+  n_cols = data.n_cols;
 
+  //verify the PCA results based on the loadings(coeff matrix) and the score
+  for(int i = 0; i < n_rows; i++)
+  {
+    for(int j = 0; j < n_cols; j++)
+    {
+      assert(fabs(coeff_actual(i, j) - coeff(i, j)) < 0.0001);
+      assert(fabs(score_actual(i, j) - score(i, j)) < 0.0001);
+    }
+  }
 
-	mat trans_data = arma::trans(data);
-	mlpack::pca::PCA p;
-	p.Apply(coeff, score, trans_data);
-
-	n_rows = data.n_rows;
-	n_cols = data.n_cols;
-
-	/*cout << "test matrix : " << endl << data;
-
-	cout << "coeff : " << endl << coeff;
-
-	cout << "score : " << endl << score;*/
-
-	for(int i = 0; i < n_rows; i++)
-	{
-		for(int j = 0; j < n_cols; j++)
-		{
-			assert(fabs(coeff_actual(i, j) - coeff(i, j)) < 0.0001);
-			assert(fabs(score_actual(i, j) - score(i, j)) < 0.0001);
-		}
-	}
-
-	return 0;
+  return 0;
 }