From aa94bf7dc2e3c025bd3197fabe2a90141d96332f Mon Sep 17 00:00:00 2001
From: AYESDIE <34449856+AYESDIE@users.noreply.github.com>
Date: Tue, 23 Oct 2018 18:04:46 +0530
Subject: [PATCH] Added support for different Interpolation and Neighbout
 Search Algorithms

Added support for different Interpolation and Neighbout Search Algorithms
---
 src/mlpack/methods/cf/cf_main.cpp       | 173 +++++++++++++++++++++++-
 src/mlpack/methods/cf/cf_model.hpp      |  10 ++
 src/mlpack/methods/cf/cf_model_impl.hpp |  36 +++--
 3 files changed, 206 insertions(+), 13 deletions(-)

diff --git a/src/mlpack/methods/cf/cf_main.cpp b/src/mlpack/methods/cf/cf_main.cpp
index 0d0741818f..ad896c122e 100644
--- a/src/mlpack/methods/cf/cf_main.cpp
+++ b/src/mlpack/methods/cf/cf_main.cpp
@@ -25,6 +25,15 @@
 #include <mlpack/methods/cf/decomposition_policies/bias_svd_method.hpp>
 #include <mlpack/methods/cf/decomposition_policies/svdplusplus_method.hpp>
 
+#include <mlpack/methods/cf/interpolation_policies/average_interpolation.hpp>
+#include <mlpack/methods/cf/interpolation_policies/regression_interpolation.hpp>
+#include <mlpack/methods/cf/interpolation_policies/similarity_interpolation.hpp>
+
+#include <mlpack/methods/cf/neighbor_search_policies/cosine_search.hpp>
+#include <mlpack/methods/cf/neighbor_search_policies/lmetric_search.hpp>
+#include <mlpack/methods/cf/neighbor_search_policies/pearson_search.hpp>
+
+
 using namespace mlpack;
 using namespace mlpack::cf;
 using namespace mlpack::amf;
@@ -125,10 +134,21 @@ PARAM_INT_IN("recommendations", "Number of recommendations to generate for each"
 
 PARAM_INT_IN("seed", "Set the random seed (0 uses std::time(NULL)).", "s", 0);
 
+//  Interpolation and Neighbour Search Algorithms
+PARAM_STRING_IN("interpolation", "Algorithm used for weight interpolation.", "i",
+                "Average");
+
+PARAM_STRING_IN("neighbour_search", "Algorithm used for neighbour search.", "f",
+                "Euclidean");
+
 void ComputeRecommendations(CFModel* cf,
                             const size_t numRecs,
                             arma::Mat<size_t>& recommendations)
 {
+
+  const string ns_algo = CLI::GetParam<string>("neighbour_search");
+  const string iw_algo = CLI::GetParam<string>("interpolation");
+
   // Reading users.
   if (CLI::HasParam("query"))
   {
@@ -142,17 +162,112 @@ void ComputeRecommendations(CFModel* cf,
 
     Log::Info << "Generating recommendations for " << users.n_elem << " users."
         << endl;
-    cf->GetRecommendations(numRecs, recommendations, users.row(0).t());
+
+    //  Making All the possible paths
+    if(ns_algo=="Cosine")
+    {
+        if(iw_algo=="Average")
+        {
+            cf->GetRecommendations<CosineSearch,AverageInterpolation>(numRecs, recommendations, users.row(0).t());
+        }
+        else if(iw_algo=="Regression")
+        {
+            cf->GetRecommendations<CosineSearch,RegressionInterpolation>(numRecs, recommendations, users.row(0).t());
+        }
+        else if(iw_algo=="Similarity")
+        {
+            cf->GetRecommendations<CosineSearch,SimilarityInterpolation>(numRecs, recommendations, users.row(0).t());
+        }
+    }
+    else if(ns_algo=="Euclidean")
+    {
+        if(iw_algo=="Average")
+        {
+            cf->GetRecommendations<EuclideanSearch,AverageInterpolation>(numRecs, recommendations, users.row(0).t());
+        }
+        else if(iw_algo=="Regression")
+        {
+            cf->GetRecommendations<EuclideanSearch,RegressionInterpolation>(numRecs, recommendations, users.row(0).t());
+        }
+        else if(iw_algo=="Similarity")
+        {
+            cf->GetRecommendations<EuclideanSearch,SimilarityInterpolation>(numRecs, recommendations, users.row(0).t());
+        }
+    }
+    else if(ns_algo=="Pearson")
+    {
+        if(iw_algo=="Average")
+        {
+            cf->GetRecommendations<PearsonSearch,AverageInterpolation>(numRecs, recommendations, users.row(0).t());
+        }
+        else if(iw_algo=="Regression")
+        {
+            cf->GetRecommendations<PearsonSearch,RegressionInterpolation>(numRecs, recommendations, users.row(0).t());
+        }
+        else if(iw_algo=="Similarity")
+        {
+            cf->GetRecommendations<PearsonSearch,SimilarityInterpolation>(numRecs, recommendations, users.row(0).t());
+        }
+    }
   }
   else
   {
     Log::Info << "Generating recommendations for all users." << endl;
-    cf->GetRecommendations(numRecs, recommendations);
+      if(ns_algo=="Cosine")
+      {
+          if(iw_algo=="Average")
+          {
+              cf->GetRecommendations<CosineSearch,AverageInterpolation>(numRecs, recommendations);
+          }
+          else if(iw_algo=="Regression")
+          {
+              cf->GetRecommendations<CosineSearch,RegressionInterpolation>(numRecs, recommendations);
+          }
+          else if(iw_algo=="Similarity")
+          {
+              cf->GetRecommendations<CosineSearch,SimilarityInterpolation>(numRecs, recommendations);
+          }
+      }
+      else if(ns_algo=="Euclidean")
+      {
+          if(iw_algo=="Average")
+          {
+              cf->GetRecommendations<EuclideanSearch,AverageInterpolation>(numRecs, recommendations);
+          }
+          else if(iw_algo=="Regression")
+          {
+              cf->GetRecommendations<EuclideanSearch,RegressionInterpolation>(numRecs, recommendations);
+          }
+          else if(iw_algo=="Similarity")
+          {
+              cf->GetRecommendations<EuclideanSearch,SimilarityInterpolation>(numRecs, recommendations);
+          }
+      }
+      else if(ns_algo=="Pearson")
+      {
+          if(iw_algo=="Average")
+          {
+              cf->GetRecommendations<PearsonSearch,AverageInterpolation>(numRecs, recommendations);
+          }
+          else if(iw_algo=="Regression")
+          {
+              cf->GetRecommendations<PearsonSearch,RegressionInterpolation>(numRecs, recommendations);
+          }
+          else if(iw_algo=="Similarity")
+          {
+              cf->GetRecommendations<PearsonSearch,SimilarityInterpolation>(numRecs, recommendations);
+          }
+      }
   }
 }
 
 void ComputeRMSE(CFModel* cf)
 {
+  //    Interpolation and Neighbour Search
+
+  const string ns_algo = CLI::GetParam<string>("neighbour_search");
+  const string iw_algo = CLI::GetParam<string>("interpolation");
+
   // Now, compute each test point.
   arma::mat testData = std::move(CLI::GetParam<arma::mat>("test"));
 
@@ -166,7 +281,52 @@ void ComputeRMSE(CFModel* cf)
 
   // Now compute the RMSE.
   arma::vec predictions;
-  cf->Predict(combinations, predictions);
+
+  if(ns_algo=="Cosine")
+    {
+        if(iw_algo=="Average")
+        {
+            cf->Predict<CosineSearch,AverageInterpolation>(combinations, predictions);
+        }
+        else if(iw_algo=="Regression")
+        {
+            cf->Predict<CosineSearch,RegressionInterpolation>(combinations, predictions);
+        }
+        else if(iw_algo=="Similarity")
+        {
+            cf->Predict<CosineSearch,SimilarityInterpolation>(combinations, predictions);
+        }
+    }
+  else if(ns_algo=="Euclidean")
+    {
+        if(iw_algo=="Average")
+        {
+            cf->Predict<EuclideanSearch,AverageInterpolation>(combinations, predictions);
+        }
+        else if(iw_algo=="Regression")
+        {
+            cf->Predict<EuclideanSearch,RegressionInterpolation>(combinations, predictions);
+        }
+        else if(iw_algo=="Similarity")
+        {
+            cf->Predict<EuclideanSearch,SimilarityInterpolation>(combinations, predictions);
+        }
+    }
+  else if(ns_algo=="Pearson")
+    {
+        if(iw_algo=="Average")
+        {
+            cf->Predict<PearsonSearch,AverageInterpolation>(combinations, predictions);
+        }
+        else if(iw_algo=="Regression")
+        {
+            cf->Predict<PearsonSearch,RegressionInterpolation>(combinations, predictions);
+        }
+        else if(iw_algo=="Similarity")
+        {
+            cf->Predict<PearsonSearch,SimilarityInterpolation>(combinations, predictions);
+        }
+    }
 
   // Compute the root of the sum of the squared errors, divide by the number of
   // points to get the RMSE.  It turns out this is just the L2-norm divided by
@@ -287,6 +447,13 @@ static void mlpackMain()
       "SVDIncompleteIncremental", "SVDCompleteIncremental", "RegSVD",
       "RandSVD", "BiasSVD", "SVDPP" }, true, "unknown algorithm");
 
+    //    Validate the interpolation and neighbour_search policy
+  RequireParamInSet<string>("interpolation", { "Average", "Regression", "Similarity" },
+      true, "unknown interpolation algorithm");
+
+  RequireParamInSet<string>("neighbour_search", { "Cosine", "Euclidean", "Pearson" },
+      true, "unknown neighbour search algorithm");
+
   ReportIgnoredParam({{ "iteration_only_termination", true }}, "min_residue");
 
   RequireParamValue<int>("recommendations", [](int x) { return x > 0; }, true,
diff --git a/src/mlpack/methods/cf/cf_model.hpp b/src/mlpack/methods/cf/cf_model.hpp
index 0b82687bbc..ae568e371d 100644
--- a/src/mlpack/methods/cf/cf_model.hpp
+++ b/src/mlpack/methods/cf/cf_model.hpp
@@ -54,6 +54,8 @@ class GetValueVisitor : public boost::static_visitor<void*>
  * PredictVisitor uses the CFType object to make predictions on the given
  * combinations of users and items.
  */
+template <typename NeighbourSearchPolicy,
+          typename InterpolationPolicy>
 class PredictVisitor : public boost::static_visitor<void>
 {
  private:
@@ -76,6 +78,8 @@ class PredictVisitor : public boost::static_visitor<void>
  * RecommendationVisitor uses the CFType object to get recommendations for the
  * given users.
  */
+template <typename NeighbourSearchPolicy,
+          typename InterpolationPolicy>
 class RecommendationVisitor : public boost::static_visitor<void>
 {
  private:
@@ -142,15 +146,21 @@ class CFModel
              const bool mit);
 
   //! Make predictions.
+  template <typename NeighbourSearchPolicy,
+            typename InterpolationPolicy>
   void Predict(const arma::Mat<size_t>& combinations,
                arma::vec& predictions);
 
   //! Compute recommendations for query users.
+  template<typename NeighbourSearchPolicy,
+           typename InterpolationPolicy>
   void GetRecommendations(const size_t numRecs,
                           arma::Mat<size_t>& recommendations,
                           const arma::Col<size_t>& users);
 
   //! Compute recommendations for all users.
+  template<typename NeighbourSearchPolicy,
+          typename InterpolationPolicy>
   void GetRecommendations(const size_t numRecs,
                           arma::Mat<size_t>& recommendations);
 
diff --git a/src/mlpack/methods/cf/cf_model_impl.hpp b/src/mlpack/methods/cf/cf_model_impl.hpp
index 2c54b13e5d..5f9f244b6c 100644
--- a/src/mlpack/methods/cf/cf_model_impl.hpp
+++ b/src/mlpack/methods/cf/cf_model_impl.hpp
@@ -34,15 +34,19 @@ void* GetValueVisitor::operator()(CFType<DecompositionPolicy>* c) const
   return (void*) c;
 }
 
-PredictVisitor::PredictVisitor(
+template <typename NeighbourSearchPolicy,
+          typename InterpolationPolicy>
+PredictVisitor<NeighbourSearchPolicy,InterpolationPolicy>::PredictVisitor(
     const arma::Mat<size_t>& combinations,
     arma::vec& predictions) :
     combinations(combinations),
     predictions(predictions)
 { }
 
+template <typename NeighbourSearchPolicy,
+          typename InterpolationPolicy>
 template<typename DecompositionPolicy>
-void PredictVisitor::operator()(CFType<DecompositionPolicy>* c) const
+void PredictVisitor<NeighbourSearchPolicy,InterpolationPolicy>::operator()(CFType<DecompositionPolicy>* c) const
 {
   if (!c)
   {
@@ -50,10 +54,13 @@ void PredictVisitor::operator()(CFType<DecompositionPolicy>* c) const
     return;
   }
 
-  c->Predict(combinations, predictions);
+  c->template Predict<NeighbourSearchPolicy,InterpolationPolicy>(combinations, predictions);
 }
 
-RecommendationVisitor::RecommendationVisitor(
+template <typename NeighbourSearchPolicy,
+          typename InterpolationPolicy>
+RecommendationVisitor<NeighbourSearchPolicy,InterpolationPolicy>
+        ::RecommendationVisitor(
     const size_t numRecs,
     arma::Mat<size_t>& recommendations,
     const arma::Col<size_t>& users,
@@ -64,8 +71,11 @@ RecommendationVisitor::RecommendationVisitor(
     usersGiven(usersGiven)
 { }
 
+template <typename NeighbourSearchPolicy,
+        typename InterpolationPolicy>
 template<typename DecompositionPolicy>
-void RecommendationVisitor::operator()(CFType<DecompositionPolicy>* c) const
+void RecommendationVisitor<NeighbourSearchPolicy,InterpolationPolicy>
+        ::operator()(CFType<DecompositionPolicy>* c) const
 {
   if (!c)
   {
@@ -74,9 +84,9 @@ void RecommendationVisitor::operator()(CFType<DecompositionPolicy>* c) const
   }
 
   if (usersGiven)
-    c->GetRecommendations(numRecs, recommendations, users);
+    c->template GetRecommendations<NeighbourSearchPolicy,InterpolationPolicy>(numRecs, recommendations, users);
   else
-    c->GetRecommendations(numRecs, recommendations);
+    c->template GetRecommendations<NeighbourSearchPolicy,InterpolationPolicy>(numRecs, recommendations);
 }
 
 CFModel::~CFModel()
@@ -103,28 +113,34 @@ void CFModel::Train(const MatType& data,
 }
 
 //! Make predictions.
+template <typename NeighbourSearchPolicy,
+          typename InterpolationPolicy>
 void CFModel::Predict(const arma::Mat<size_t>& combinations,
                       arma::vec& predictions)
 {
-  PredictVisitor predict(combinations, predictions);
+  PredictVisitor<NeighbourSearchPolicy,InterpolationPolicy> predict(combinations, predictions);
   boost::apply_visitor(predict, cf);
 }
 
 //! Compute recommendations for queried users.
+template<typename NeighbourSearchPolicy,
+        typename InterpolationPolicy>
 void CFModel::GetRecommendations(const size_t numRecs,
                                  arma::Mat<size_t>& recommendations,
                                  const arma::Col<size_t>& users)
 {
-  RecommendationVisitor recommendation(numRecs, recommendations, users, true);
+  RecommendationVisitor<NeighbourSearchPolicy,InterpolationPolicy> recommendation(numRecs, recommendations, users, true);
   boost::apply_visitor(recommendation, cf);
 }
 
 //! Compute recommendations for all users.
+template<typename NeighbourSearchPolicy,
+        typename InterpolationPolicy>
 void CFModel::GetRecommendations(const size_t numRecs,
                                  arma::Mat<size_t>& recommendations)
 {
   arma::Col<size_t> users;
-  RecommendationVisitor recommendation(numRecs, recommendations, users, false);
+  RecommendationVisitor<NeighbourSearchPolicy,InterpolationPolicy> recommendation(numRecs, recommendations, users, false);
   boost::apply_visitor(recommendation, cf);
 }