first parallel optimizer

src/Makevars

@@ -1,6 +1,12 @@
-PKG_CFLAGS = $(SHLIB_OPENMP_CFLAGS)
+PKG_LIBS = $(LAPACK_LIBS) $(BLAS_LIBS) $(FLIBS) $(SHLIB_OPENMP_CXXFLAGS)
 CXX_STD = CXX11
-PKG_LIBS = $(SHLIB_OPENMP_CFLAGS) $(LAPACK_LIBS) $(BLAS_LIBS) $(FLIBS)
+
+PKG_CXXFLAGS=$(SHLIB_OPENMP_CXXFLAGS)
+
+SHLIB_OPENMP_CFLAGS=$(SHLIB_OPENMP_CFLAGS)
+SHLIB_OPENMP_CXXFLAGS=$(SHLIB_OPENMP_CXXFLAGS)
+SHLIB_OPENMP_FCFLAGS=$(SHLIB_OPENMP_FCFLAGS)
+SHLIB_OPENMP_FFLAGS=$(SHLIB_OPENMP_FFLAGS)
 
 # strip debug symbols for smaller Linux binaries
 strippedLib: $(SHLIB)

src/Makevars.win

@@ -1,3 +1,9 @@
-PKG_CFLAGS = $(SHLIB_OPENMP_CFLAGS)
+PKG_LIBS = $(LAPACK_LIBS) $(BLAS_LIBS) $(FLIBS) $(SHLIB_OPENMP_CXXFLAGS)
 CXX_STD = CXX11
-PKG_LIBS = $(SHLIB_OPENMP_CFLAGS) $(LAPACK_LIBS) $(BLAS_LIBS) $(FLIBS)
+
+PKG_CXXFLAGS=$(SHLIB_OPENMP_CXXFLAGS)
+
+SHLIB_OPENMP_CFLAGS=$(SHLIB_OPENMP_CFLAGS)
+SHLIB_OPENMP_CXXFLAGS=$(SHLIB_OPENMP_CXXFLAGS)
+SHLIB_OPENMP_FCFLAGS=$(SHLIB_OPENMP_FCFLAGS)
+SHLIB_OPENMP_FFLAGS=$(SHLIB_OPENMP_FFLAGS)

src/compboost_modules.cpp

@@ -2097,7 +2097,14 @@ class OptimizerWrapper
 class OptimizerCoordinateDescent : public OptimizerWrapper
 {
 public:
-  OptimizerCoordinateDescent () { sh_ptr_optimizer = std::make_shared<optimizer::OptimizerCoordinateDescent>(); }
+
+  OptimizerCoordinateDescent () {
+    sh_ptr_optimizer = std::make_shared<optimizer::OptimizerCoordinateDescent>();
+  }
+
+  OptimizerCoordinateDescent (unsigned int num_threads) {
+    sh_ptr_optimizer = std::make_shared<optimizer::OptimizerCoordinateDescent>(num_threads);
+  }
 };
 
 //' Coordinate Descent with line search
@@ -2147,6 +2154,7 @@ RCPP_MODULE(optimizer_module)
   class_<OptimizerCoordinateDescent> ("OptimizerCoordinateDescent")
     .derives<OptimizerWrapper> ("Optimizer")
     .constructor ()
+    .constructor <unsigned int> ()
   ;
 
   class_<OptimizerCoordinateDescentLineSearch> ("OptimizerCoordinateDescentLineSearch")

src/optimizer.cpp

@@ -43,9 +43,72 @@ OptimizerCoordinateDescent::OptimizerCoordinateDescent () {
   step_sizes.assign(1, 1.0);
 }
 
+OptimizerCoordinateDescent::OptimizerCoordinateDescent (const unsigned int& num_threads)
+  : num_threads (num_threads) {
+  // Initialize step size vector as scalar:
+  step_sizes.assign(1, 1.0);
+}
+
 std::shared_ptr<blearner::Baselearner> OptimizerCoordinateDescent::findBestBaselearner (const std::string& iteration_id,
   std::shared_ptr<response::Response> sh_ptr_response, const blearner_factory_map& my_blearner_factory_map) const
 {
+  std::map<double, std::shared_ptr<blearner::Baselearner>> best_blearner_map;
+
+  #pragma omp parallel num_threads(num_threads) default(none) shared(iteration_id, sh_ptr_response, my_blearner_factory_map, best_blearner_map)
+  {
+    // private per core:
+    double ssq_best = std::numeric_limits<double>::infinity();
+    double ssq_temp;
+    std::shared_ptr<blearner::Baselearner> blearner_temp;
+    std::shared_ptr<blearner::Baselearner> blearner_best;
+
+    #pragma omp for
+    for (unsigned int i = 0; i < my_blearner_factory_map.size(); i++) {
+
+      // increment iterator to "index map elements by index" (https://stackoverflow.com/questions/8848870/use-openmp-in-iterating-over-a-map):
+      auto it_factory_pair = my_blearner_factory_map.begin();
+      std::advance(it_factory_pair, i);
+
+      // Paste string identifier for new base-learner:
+      std::string id = "(" + iteration_id + ") " + it_factory_pair->second->getBaselearnerType();
+
+      // Create new base-learner out of the actual factory (just the
+      // pointer is overwritten):
+      blearner_temp = it_factory_pair->second->createBaselearner(id);
+      blearner_temp->train(sh_ptr_response->getPseudoResiduals());
+      ssq_temp = helper::calculateSumOfSquaredError(sh_ptr_response->getPseudoResiduals(), blearner_temp->predict());
+
+      // Check if SSE of new temporary base-learner is smaller then SSE of the best
+      // base-learner. If so, assign the temporary base-learner with the best
+      // base-learner (This is always triggered within the first iteration since
+      // ssq_best is declared as infinity):
+      if (ssq_temp < ssq_best) {
+        ssq_best = ssq_temp;
+        // // Deep copy since the temporary base-learner is deleted every time which
+        // // will also deletes the data for the best base-learner if we don't copy
+        // // the whole data of the object:
+        // blearner_best = blearner_temp->clone();
+        blearner_best = blearner_temp;
+      }
+    }
+
+    #pragma omp critical
+    {
+      // return best blearner per core with corresponding ssq to master:
+      best_blearner_map[ssq_best] = blearner_best;
+    }
+  }
+  if (best_blearner_map.size() == 1) {
+    return best_blearner_map.begin()->second;
+  } else {
+    auto it = min_element(best_blearner_map.begin(), best_blearner_map.end(),
+      [](decltype(best_blearner_map)::value_type& l, decltype(best_blearner_map)::value_type& r) -> bool { return l.first < r.first; });
+    return it->second;
+  }
+
+  /* ****************************************************************************************
+   * OLD SEQUENTIAL LOOP:
+   *
   double ssq_temp;
   double ssq_best = std::numeric_limits<double>::infinity();
 
@@ -76,7 +139,8 @@ std::shared_ptr<blearner::Baselearner> OptimizerCoordinateDescent::findBestBasel
       blearner_best = blearner_temp;
     }
   }
-  return blearner_best;
+  **************************************************************************************** */
+  // return blearner_best;
 }
 
 void OptimizerCoordinateDescent::calculateStepSize (std::shared_ptr<loss::Loss> sh_ptr_loss, std::shared_ptr<response::Response> sh_ptr_response,

src/optimizer.h

@@ -34,6 +34,10 @@
 #include "line_search.h"
 #include "helper.h"
 
+#ifdef _OPENMP
+#include <omp.h>
+#endif
+
 namespace optimizer {
 
 // -------------------------------------------------------------------------- //
@@ -70,8 +74,11 @@ class Optimizer
 class OptimizerCoordinateDescent : public Optimizer
 {
   public:
-    // No special initialization necessary:
+
+    unsigned int num_threads = 1;
+
     OptimizerCoordinateDescent ();
+    OptimizerCoordinateDescent (const unsigned int&);
 
     std::shared_ptr<blearner::Baselearner> findBestBaselearner (const std::string&, std::shared_ptr<response::Response>,
       const blearner_factory_map&) const;