From 04d2b44ba34a811e1fab0b90d38ebd06cd918c52 Mon Sep 17 00:00:00 2001
From: Francois Fleuret <francois@fleuret.org>
Date: Wed, 27 Mar 2013 21:11:48 +0100
Subject: [PATCH] Cosmetics.

---
 clusterer.cc | 26 ++++++++++++--------------
 clusterer.h  |  1 +
 2 files changed, 13 insertions(+), 14 deletions(-)

diff --git a/clusterer.cc b/clusterer.cc
index 3c33f3c..04a9af4 100644
--- a/clusterer.cc
+++ b/clusterer.cc
@@ -35,6 +35,8 @@ Clusterer::~Clusterer() {
 }
 
 scalar_t Clusterer::distance_to_centroid(scalar_t *x, int k) {
+  // We take the variance into account + the normalization term. This
+  // is between k-mean and EM with a diagonal covariance
   scalar_t dist = 0;
   for(int d = 0; d < _dim; d++) {
     dist += sq(_cluster_means[k][d] - x[d]) / (2 * _cluster_var[k][d]);
@@ -185,7 +187,8 @@ scalar_t Clusterer::uninformative_lp_cluster_association(int nb_points, scalar_t
   // association coefficients is equal to 1.0
 
   for(int n = 1; n <= nb_points; n++) {
-    glp_set_row_bnds(lp, n, GLP_FX, 1.0, 1.0);
+    int row = n;
+    glp_set_row_bnds(lp, row, GLP_FX, 1.0, 1.0);
   }
 
   // (B) For each pair cluster/class, the sum of the association
@@ -207,24 +210,17 @@ scalar_t Clusterer::uninformative_lp_cluster_association(int nb_points, scalar_t
 
   for(int k = 1; k <= _nb_clusters; k++) {
     for(int n = 1; n <= nb_points; n++) {
-      int r = n + nb_points * (k - 1);
-
-      // scalar_t dist = 0;
-
-      // for(int d = 0; d < _dim; d++) {
-        // dist += sq(_cluster_means[k-1][d] - points[n-1][d]) / (2 * _cluster_var[k-1][d]);
-        // dist += 0.5 * log(_cluster_var[k-1][d]);
-      // }
+      int col = n + nb_points * (k - 1);
 
       // The LP weight on this association coefficient for the global
       // loss is the normalized distance of that sample to the
       // centroid of that cluster
 
-      glp_set_obj_coef(lp, r, distance_to_centroid(points[n-1], k-1));
+      glp_set_obj_coef(lp, col, distance_to_centroid(points[n-1], k-1));
 
-      // And this association coefficient is in [0,1]
+      // And all the association coefficient is in [0,1]
 
-      glp_set_col_bnds(lp, r, GLP_DB, 0.0, 1.0);
+      glp_set_col_bnds(lp, col, GLP_DB, 0.0, 1.0);
     }
   }
 
@@ -311,11 +307,13 @@ void Clusterer::update_clusters(int nb_points, scalar_t **points, scalar_t **gam
 
     for(int d = 0; d < _dim; d++) {
       if(sum_gamma >= 2) {
-        _cluster_var[k][d] = (_cluster_var[k][d] - sq(_cluster_means[k][d]) / sum_gamma) / (sum_gamma - 1);
+        _cluster_var[k][d] =
+          (_cluster_var[k][d] - sq(_cluster_means[k][d]) / sum_gamma) / (sum_gamma - 1);
+        _cluster_var[k][d] = max(scalar_t(min_cluster_variance), _cluster_var[k][d]);
       } else {
         _cluster_var[k][d] = 1;
       }
-      _cluster_var[k][d] = max(0.01, _cluster_var[k][d]);
+
       _cluster_means[k][d] /= sum_gamma;
     }
   }
diff --git a/clusterer.h b/clusterer.h
index f73d6c7..ad0c58f 100644
--- a/clusterer.h
+++ b/clusterer.h
@@ -38,6 +38,7 @@ public:
 
   const static int max_nb_iterations = 10;
   const static scalar_t min_iteration_improvement = 0.999;
+  const static scalar_t min_cluster_variance = 0.01f;
 
   int _nb_clusters;
   int _dim;
-- 
2.39.5