Distributed kde test passes.

fastlib/trunk/contrib/dongryel/thesis_research/core/gnp/distributed_dualtree_dfs_dev.h

@@ -30,8 +30,6 @@ void core::gnp::DistributedDualtreeDfs<DistributedProblemType>::AllReduce_(
   self_argument.Init(problem_->global());
   self_problem.Init(self_argument);
   self_engine.Init(self_problem);
-  self_problem.global().set_effective_num_reference_points(
-    problem_->global().effective_num_reference_points());
   self_engine.Compute(metric, query_results);
   world_->barrier();
 
@@ -103,8 +101,6 @@ void core::gnp::DistributedDualtreeDfs<DistributedProblemType>::AllReduce_(
         query_table_->local_table(),
         problem_->global());
       sub_problem.Init(sub_argument);
-      sub_problem.global().set_effective_num_reference_points(
-        problem_->global().effective_num_reference_points());
       sub_engine.Init(sub_problem);
       sub_engine.Compute(metric, query_results, false);
     }

fastlib/trunk/contrib/dongryel/thesis_research/mlpack/distributed_kde/distributed_kde_dev.h

@@ -86,7 +86,8 @@ void DistributedKde<DistributedTableType>::Init(
 
   // Declare the global constants.
   global_.Init(
-    reference_table_, query_table_, arguments_in.bandwidth_, is_monochromatic_,
+    reference_table_, query_table_, reference_table_->n_entries(),
+    arguments_in.bandwidth_, is_monochromatic_,
     arguments_in.relative_error_, arguments_in.probability_,
     arguments_in.kernel_, false);
   global_.set_effective_num_reference_points(
@@ -119,11 +120,11 @@ bool DistributedKde<DistributedTableType>::ConstructBoostVariableMap_(
     "the leave-one-out density at each reference point."
   )(
     "random_generate_n_attributes",
-    boost::program_options::value<int>()->default_value(3),
+    boost::program_options::value<int>(),
     "Generate the datasets on the fly of the specified dimension."
   )(
     "random_generate_n_entries",
-    boost::program_options::value<int>()->default_value(20),
+    boost::program_options::value<int>(),
     "Generate the datasets on the fly of the specified number of points."
   )(
     "densities_out",

fastlib/trunk/contrib/dongryel/thesis_research/mlpack/kde/kde_arguments.h

@@ -23,6 +23,8 @@ class KdeArguments {
 
     TableType *query_table_;
 
+    double effective_num_reference_points_;
+
     double bandwidth_;
 
     double relative_error_;
@@ -35,6 +37,8 @@ class KdeArguments {
 
     bool tables_are_aliased_;
 
+    bool normalize_densities_;
+
   public:
 
     template<typename GlobalType>
@@ -43,11 +47,14 @@ class KdeArguments {
       GlobalType &global_in) {
       reference_table_ = reference_table_in;
       query_table_ = query_table_in;
+      effective_num_reference_points_ =
+        global_in.effective_num_reference_points();
       bandwidth_ = global_in.bandwidth();
       relative_error_ = global_in.relative_error();
       probability_ = global_in.probability();
       kernel_ = global_in.kernel().name();
       tables_are_aliased_ = true;
+      normalize_densities_ = global_in.normalize_densities();
     }
 
     template<typename GlobalType>
@@ -56,23 +63,28 @@ class KdeArguments {
       if(reference_table_ != query_table_) {
         query_table_ = global_in.query_table()->local_table();
       }
+      effective_num_reference_points_ =
+        global_in.effective_num_reference_points();
       bandwidth_ = global_in.bandwidth();
       relative_error_ = global_in.relative_error();
       probability_ = global_in.probability();
       kernel_ = global_in.kernel().name();
       tables_are_aliased_ = true;
+      normalize_densities_ = global_in.normalize_densities();
     }
 
     KdeArguments() {
       leaf_size_ = 0;
       reference_table_ = NULL;
       query_table_ = NULL;
+      effective_num_reference_points_ = 0.0;
       bandwidth_ = 0.0;
       relative_error_ = 0.0;
       probability_ = 0.0;
       kernel_ = "";
       metric_ = NULL;
       tables_are_aliased_ = false;
+      normalize_densities_ = true;
     }
 
     ~KdeArguments() {

fastlib/trunk/contrib/dongryel/thesis_research/mlpack/kde/kde_dev.h

@@ -60,9 +60,11 @@ void mlpack::kde::Kde<TableType>::Init(
 
   // Declare the global constants.
   global_.Init(
-    reference_table_, query_table_, arguments_in.bandwidth_, is_monochromatic_,
+    reference_table_, query_table_,
+    arguments_in.effective_num_reference_points_,
+    arguments_in.bandwidth_, is_monochromatic_,
     arguments_in.relative_error_, arguments_in.probability_,
-    arguments_in.kernel_);
+    arguments_in.kernel_, arguments_in.normalize_densities_);
 }
 
 template<typename TableType>
@@ -221,9 +223,13 @@ void mlpack::kde::Kde<TableType>::ParseArguments(
     arguments_out->query_table_->IndexData(
       *(arguments_out->metric_), arguments_out->leaf_size_);
     std::cout << "Finished building the query tree.\n";
+    arguments_out->effective_num_reference_points_ =
+      arguments_out->reference_table_->n_entries();
   }
   else {
     arguments_out->query_table_ = arguments_out->reference_table_;
+    arguments_out->effective_num_reference_points_ =
+      arguments_out->reference_table_->n_entries() - 1;
   }
 
   // Parse the bandwidth.

fastlib/trunk/contrib/dongryel/thesis_research/mlpack/kde/kde_dualtree.h

@@ -144,16 +144,6 @@ class KdeGlobal {
       return effective_num_reference_points_;
     }
 
-    void set_effective_num_reference_points(
-      double effective_num_reference_points_in) {
-
-      effective_num_reference_points_ = effective_num_reference_points_in;
-      mult_const_ = 1.0 /
-                    (kernel_->CalcNormConstant(
-                       reference_table_->n_attributes()) *
-                     ((double) effective_num_reference_points_));
-    }
-
     template<typename DistributedTableType>
     void set_effective_num_reference_points(
       boost::mpi::communicator &comm,
@@ -164,10 +154,13 @@ class KdeGlobal {
       for(int i = 0; i < comm.size(); i++) {
         total_sum += reference_table_in->local_n_entries(i);
       }
-      double effective_num_reference_points_in =
+      effective_num_reference_points_ =
         (reference_table_in == query_table_in) ?
         (total_sum - 1.0) : total_sum;
-      set_effective_num_reference_points(effective_num_reference_points_in);
+      mult_const_ = 1.0 /
+                    (kernel_->CalcNormConstant(
+                       reference_table_in->n_attributes()) *
+                     ((double) effective_num_reference_points_));
     }
 
     ~KdeGlobal() {
@@ -228,15 +221,13 @@ class KdeGlobal {
     void Init(
       TableType *reference_table_in,
       TableType *query_table_in,
+      int effective_num_reference_points_in,
       double bandwidth_in, const bool is_monochromatic,
       double relative_error_in, double probability_in,
       const std::string &kernel_type_in,
       bool normalize_densities_in = true) {
 
-      effective_num_reference_points_ =
-        (is_monochromatic) ?
-        (reference_table_in->n_entries() - 1) :
-        reference_table_in->n_entries();
+      effective_num_reference_points_ = effective_num_reference_points_in;
 
       if(kernel_type_in == "gaussian") {
         kernel_ = new core::metric_kernels::GaussianKernel();