Changeset 813

Timestamp:

Mar 16, 2007, 8:30:02 PM (17 years ago)

Author:

Peter

Message:

Predict in NBC. Fixes #57

Location:

trunk

Files:

• test/Makefile.am (modified) (2 diffs)
• test/nbc_test.cc (added)
• yat/classifier/NBC.cc (modified) (2 diffs)
• yat/classifier/NBC.h (modified) (1 diff)
• yat/utility/matrix.cc (modified) (2 diffs)

trunk/test/Makefile.am

@@ -30 +30 @@
   ensemble_test feature_selection_test fileutil_test inputranker_test \
   kernel_test kernel_lookup_test matrix_test matrix_lookup_test       \
+  nbc_test \
   ncc_test nni_test pca_test regression_test rnd_test roc_test \
   score_test      \
@@ -57 +58 @@
 matrix_test_SOURCES = matrix_test.cc
 matrix_lookup_test_SOURCES = matrix_lookup_test.cc
+nbc_test_SOURCES = nbc_test.cc
 ncc_test_SOURCES = ncc_test.cc
 nni_test_SOURCES = nni_test.cc

trunk/yat/classifier/NBC.cc

@@ -90 +90 @@
           aver[target_(j)].add(data_(i,j),1.0);
       }
-      for (size_t j=0; target_.nof_classes(); ++j){
+      assert(centroids_.columns()==target_.nof_classes());
+      for (size_t j=0; j<target_.nof_classes(); ++j){
+        assert(i<centroids_.rows());
+        assert(j<centroids_.columns());
         centroids_(i,j) = aver[j].mean();
+        assert(i<sigma2_.rows());
+        assert(j<sigma2_.columns());
         sigma2_(i,j) = aver[j].variance();
       }
@@ -104 +109 @@
   {   
     assert(data_.rows()==x.rows());
+    assert(x.rows()==sigma2_.rows());
+    assert(x.rows()==centroids_.rows());
+
+    const MatrixLookupWeighted* w = 
+      dynamic_cast<const MatrixLookupWeighted*>(&x);
 
     // each row in prediction corresponds to a sample label (class)
     prediction.resize(centroids_.columns(), x.columns(), 0);
     // first calculate -lnP = sum sigma_i + (x_i-m_i)^2/2sigma_i^2
-    for (size_t label=0; label<prediction.columns(); ++label) {
+    for (size_t label=0; label<centroids_.columns(); ++label) {
       double sum_ln_sigma=0;
-      for (size_t i=0; i<x.rows(); ++i) 
+      assert(label<sigma2_.columns());
+      for (size_t i=0; i<x.rows(); ++i) {
+        assert(i<sigma2_.rows());
         sum_ln_sigma += std::log(sigma2_(i, label));
+      }
       sum_ln_sigma /= 2; // taking sum of log(sigma) not sigma2
       for (size_t sample=0; sample<prediction.rows(); ++sample) {
         for (size_t i=0; i<x.rows(); ++i) {
-          prediction(label, sample) += 
-            std::pow(x(i, label)-centroids_(i, label),2)/sigma2_(i, label);
+          // weighted calculation
+          if (w){
+            // taking care of NaN
+            if (w->weight(i, label)){
+            prediction(label, sample) += w->weight(i, label)*
+              std::pow(w->data(i, label)-centroids_(i, label),2)/
+              sigma2_(i, label);
+            }
+          }
+          // no weights
+          else {
+            prediction(label, sample) += 
+              std::pow(x(i, label)-centroids_(i, label),2)/sigma2_(i, label);
+          }
         }
       }

trunk/yat/classifier/NBC.h

@@ -83 +83 @@
     
     /**
-       For each sample, calculate the probabilities the sample belong
-       to the corresponding class.
+       Each sample (column) in \a data is predicted and predictions
+       are returned in the corresponding column in passed \a res. Each
+       row in \a res corresponds to a class. The prediction is the
+       estimated probability that sample belong to class \f$ j \f$
+
+       \f$ P_j = \frac{1}{Z}\prod_i{\frac{1}{\sigma_i}}
+       \exp(\frac{w_i(x_i-\mu_i)^2}{\sigma_i^2})\f$, where \f$ \mu_i
+       \f$ and \f$ \sigma_i^2 \f$ are the estimated mean and variance,
+       respectively. If \a data is a MatrixLookup is equivalent to
+       using all weight equal to unity.
     */
     void predict(const DataLookup2D& data, utility::matrix& res) const;

trunk/yat/utility/matrix.cc

@@ -379 +379 @@
   {
     assert(m_);
+    assert(row<rows());
+    assert(column<columns());
     double* d=gsl_matrix_ptr(m_, row, column);
     if (!d)
@@ -388 +390 @@
   const double& matrix::operator()(size_t row, size_t column) const
   {
+    assert(row<rows());
+    assert(column<columns());
     const double* d=gsl_matrix_const_ptr(proxy_m_, row, column);
     if (!d)