X-Git-Url: https://www.fleuret.org/cgi-bin/gitweb/gitweb.cgi?p=clueless-kmeans.git;a=blobdiff_plain;f=clusterer.cc;h=965b7bacd6fe9a9fb025dc68ab886c4a1dae5d1e;hp=02a8c8b990c4cc115c5dc23be7bb2bfd3c5a81e5;hb=HEAD;hpb=513062971a6c930bd6bb1ad5a8288d326e22bb25

diff --git a/clusterer.cc b/clusterer.cc
index 02a8c8b..965b7ba 100644
--- a/clusterer.cc
+++ b/clusterer.cc
@@ -1,17 +1,17 @@
 /*
- *  clueless-kmean is a variant of k-mean which enforces balanced
+ *  clueless-kmeans is a variant of k-means which enforces balanced
  *  distribution of classes in every cluster
  *
  *  Copyright (c) 2013 Idiap Research Institute, http://www.idiap.ch/
  *  Written by Francois Fleuret <francois.fleuret@idiap.ch>
  *
- *  This file is part of clueless-kmean.
+ *  This file is part of clueless-kmeans.
  *
- *  clueless-kmean is free software: you can redistribute it and/or
+ *  clueless-kmeans is free software: you can redistribute it and/or
  *  modify it under the terms of the GNU General Public License
  *  version 3 as published by the Free Software Foundation.
  *
- *  clueless-kmean is distributed in the hope that it will be useful,
+ *  clueless-kmeans is distributed in the hope that it will be useful,
  *  but WITHOUT ANY WARRANTY; without even the implied warranty of
  *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  *  General Public License for more details.
@@ -35,8 +35,6 @@ 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]);
@@ -156,7 +154,8 @@ scalar_t Clusterer::baseline_lp_cluster_association(int nb_points, scalar_t **po
 
 scalar_t Clusterer::uninformative_lp_cluster_association(int nb_points, scalar_t **points,
                                                          int nb_classes, int *labels,
-                                                         scalar_t **gamma)  {
+                                                         scalar_t **gamma,
+                                                         int absolute_proportion)  {
   // N points
   // K clusters
   // dist(n,k) distance of samples n to cluster k
@@ -172,8 +171,8 @@ scalar_t Clusterer::uninformative_lp_cluster_association(int nb_points, scalar_t
   // under
   //
   // (A) \forall n, k, \gamma(n, k) >= 0
-  // (B) \forall n, \sum_k \gamma(n,k) = 1
-  // (C) \forall k, \sum_n \gamma(n,k) = N/K
+  // (B) \forall n, \sum_k \gamma(n, k) = 1
+  // (C) \forall k, \sum_n \gamma(n, k) = N/K
 
   glp_prob *lp;
 
@@ -182,7 +181,13 @@ scalar_t Clusterer::uninformative_lp_cluster_association(int nb_points, scalar_t
 
   // ** GLPK USES INDEXES STARTING AT 1, NOT 0. **
 
-  int nb_coeffs = nb_points * _nb_clusters + nb_points * _nb_clusters;
+  int nb_coeffs;
+
+  if(absolute_proportion) {
+    nb_coeffs = nb_points * _nb_clusters + nb_points * _nb_clusters;
+  } else {
+    nb_coeffs = nb_points * _nb_clusters + nb_points * nb_classes * _nb_clusters;
+  }
 
   int *coeff_row = new int[nb_coeffs + 1];
   int *coeff_col = new int[nb_coeffs + 1];
@@ -208,22 +213,25 @@ scalar_t Clusterer::uninformative_lp_cluster_association(int nb_points, scalar_t
 
   glp_add_cols(lp, nb_points * _nb_clusters);
 
+  // The column for gamma[n][k] point 1<=n<=nb_points and cluster
+  // 1<=k<=_nb_clusters is nb_points * (k - 1) + n;
+
   // The constraints (A) will be expressed by putting directly bounds
   // on the variables (i.e. one per column). So we need one row per
   // (B) constraint, and one per (C) constraint.
 
   glp_add_rows(lp, nb_points + _nb_clusters * nb_classes);
 
-  // First, we set the weights for the objective function, and the
-  // constraint on the individual gammas
+  // First, we set the weights for the objective function, and the (A)
+  // constraints on the individual gammas
 
   for(int k = 1; k <= _nb_clusters; k++) {
     for(int n = 1; n <= nb_points; n++) {
       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
+      // The LP weight on the gammas for the global loss is the
+      // normalized distance of that sample to the centroid of that
+      // cluster
 
       glp_set_obj_coef(lp, col, distance_to_centroid(points[n-1], k-1));
 
@@ -234,8 +242,8 @@ scalar_t Clusterer::uninformative_lp_cluster_association(int nb_points, scalar_t
     }
   }
 
-  // The (B) constraints: for each point, the sum of its association
-  // coefficients is equal to 1.0
+  // The (B) constraints: for each point, the sum of its gamma is
+  // equal to 1.0
 
   for(int n = 1; n <= nb_points; n++) {
     int row = n;
@@ -249,20 +257,36 @@ scalar_t Clusterer::uninformative_lp_cluster_association(int nb_points, scalar_t
   }
 
   // The (C) constraints: For each pair cluster/class, the sum of the
-  // association coefficient to this cluster for this class is equal
-  // to the number of sample of that class, divided by the number of
-  // clusters
+  // gammas for this cluster and this class is equal to the number of
+  // sample of that class, divided by the number of clusters
 
-  for(int k = 1; k <= _nb_clusters; k++) {
-    for(int c = 1; c <= nb_classes; c++) {
-      int row = nb_points + (k - 1) * nb_classes + c;
-      scalar_t tau = nb_samples_per_class[c-1] / scalar_t(_nb_clusters);
-      glp_set_row_bnds(lp, row, GLP_FX, tau, tau);
-      for(int n = 1; n <= nb_points; n++) {
-        if(labels[n-1] == c - 1) {
+  if(absolute_proportion) {
+    for(int k = 1; k <= _nb_clusters; k++) {
+      for(int c = 1; c <= nb_classes; c++) {
+        int row = nb_points + (k - 1) * nb_classes + c;
+        scalar_t tau = nb_samples_per_class[c-1] / scalar_t(_nb_clusters);
+        glp_set_row_bnds(lp, row, GLP_FX, tau, tau);
+        for(int n = 1; n <= nb_points; n++) {
+          if(labels[n-1] == c - 1) {
+            coeff_row[n_coeff] = row;
+            coeff_col[n_coeff] = (k-1)  * nb_points + n;
+            coeff_wgt[n_coeff] = 1.0;
+            n_coeff++;
+          }
+        }
+      }
+    }
+  } else {
+    for(int k = 1; k <= _nb_clusters; k++) {
+      for(int c = 1; c <= nb_classes; c++) {
+        int row = nb_points + (k - 1) * nb_classes + c;
+        glp_set_row_bnds(lp, row, GLP_FX, 0.0, 0.0);
+        for(int n = 1; n <= nb_points; n++) {
           coeff_row[n_coeff] = row;
           coeff_col[n_coeff] = (k-1)  * nb_points + n;
-          coeff_wgt[n_coeff] = 1.0;
+          coeff_wgt[n_coeff] =
+            (labels[n-1] == c - 1 ? 1.0 : 0.0)
+            - scalar_t(nb_samples_per_class[c-1]) / scalar_t(nb_points);
           n_coeff++;
         }
       }
@@ -357,18 +381,23 @@ void Clusterer::train(int mode,
     switch(mode) {
 
     case STANDARD_ASSOCIATION:
-    total_distance =
-      baseline_cluster_association(nb_points, points, nb_classes, labels, gammas);
+      total_distance =
+        baseline_cluster_association(nb_points, points, nb_classes, labels, gammas);
       break;
 
     case STANDARD_LP_ASSOCIATION:
-    total_distance =
-      baseline_lp_cluster_association(nb_points, points, nb_classes, labels, gammas);
+      total_distance =
+        baseline_lp_cluster_association(nb_points, points, nb_classes, labels, gammas);
       break;
 
     case UNINFORMATIVE_LP_ASSOCIATION:
-    total_distance =
-      uninformative_lp_cluster_association(nb_points, points, nb_classes, labels, gammas);
+      total_distance =
+        uninformative_lp_cluster_association(nb_points, points, nb_classes, labels, gammas, 0);
+      break;
+
+    case UNINFORMATIVE_LP_ASSOCIATION_ABSOLUTE:
+      total_distance =
+        uninformative_lp_cluster_association(nb_points, points, nb_classes, labels, gammas, 1);
       break;
 
     default: