From 4c65bb36d7e23733e3b319b24795f79443649cdf Mon Sep 17 00:00:00 2001
From: novak_99 <cooljohn312@gmail.com>
Date: Fri, 28 May 2021 19:26:10 -0700
Subject: [PATCH] "Vectorized" the implementation of SGD for Log & Lin Reg

---
 MLPP/LinReg/LinReg.cpp | 20 ++++++--------------
 MLPP/LogReg/LogReg.cpp | 20 ++++++--------------
 main.cpp               | 20 ++++++++++----------
 3 files changed, 22 insertions(+), 38 deletions(-)

diff --git a/MLPP/LinReg/LinReg.cpp b/MLPP/LinReg/LinReg.cpp
index 5bd4b03..d59c87e 100644
--- a/MLPP/LinReg/LinReg.cpp
+++ b/MLPP/LinReg/LinReg.cpp
@@ -65,6 +65,7 @@ namespace MLPP{
     }
 
     void LinReg::SGD(double learning_rate, int max_epoch, bool UI){
+        LinAlg alg;
         Reg regularization;
         Utilities util;
         double cost_prev = 0;
@@ -79,24 +80,15 @@ namespace MLPP{
             double y_hat = Evaluate(inputSet[outputIndex]);
             cost_prev = Cost({y_hat}, {outputSet[outputIndex]});
 
+            double error = y_hat - outputSet[outputIndex];
 
-            for(int i = 0; i < k; i++){
-                    
-                // Calculating the weight gradients
-                
-                double w_gradient = (y_hat - outputSet[outputIndex]) * inputSet[outputIndex][i];
-                    
-
-                // Weight updation
-                weights[i] -= learning_rate * w_gradient;
-            }
+            // Weight updation
+            weights = alg.subtraction(weights, alg.scalarMultiply(learning_rate * error, inputSet[outputIndex]));
             weights = regularization.regWeights(weights, lambda, alpha, reg);
             
-            // Calculating the bias gradients
-            double b_gradient = (y_hat - outputSet[outputIndex]);
-            
             // Bias updation
-            bias -= learning_rate * b_gradient;
+            bias -= learning_rate * error;
+
             y_hat = Evaluate({inputSet[outputIndex]});
                 
             if(UI) { 
diff --git a/MLPP/LogReg/LogReg.cpp b/MLPP/LogReg/LogReg.cpp
index 9c7d8dd..80a67e0 100644
--- a/MLPP/LogReg/LogReg.cpp
+++ b/MLPP/LogReg/LogReg.cpp
@@ -92,6 +92,7 @@ namespace MLPP{
     }
 
     void LogReg::SGD(double learning_rate, int max_epoch, bool UI){
+        LinAlg alg;
         Reg regularization;
         Utilities util;
         double cost_prev = 0;
@@ -106,24 +107,15 @@ namespace MLPP{
             double y_hat = Evaluate(inputSet[outputIndex]);
             cost_prev = Cost({y_hat}, {outputSet[outputIndex]});
 
+            double error = y_hat - outputSet[outputIndex];
 
-            for(int i = 0; i < k; i++){
-                    
-                // Calculating the weight gradients
-                
-                double w_gradient = (y_hat - outputSet[outputIndex]) * inputSet[outputIndex][i];
-                    
-
-                // Weight updation
-                weights[i] -= learning_rate * w_gradient;
-            }
+            // Weight updation
+            weights = alg.subtraction(weights, alg.scalarMultiply(learning_rate * error, inputSet[outputIndex]));
             weights = regularization.regWeights(weights, lambda, alpha, reg);
             
-            // Calculating the bias gradients
-            double b_gradient = (y_hat - outputSet[outputIndex]);
-            
             // Bias updation
-            bias -= learning_rate * b_gradient;
+            bias -= learning_rate * error;
+
             y_hat = Evaluate({inputSet[outputIndex]});
                 
             if(UI) { 
diff --git a/main.cpp b/main.cpp
index bd1a68a..6198bd6 100644
--- a/main.cpp
+++ b/main.cpp
@@ -131,26 +131,26 @@ int main() {
     // UniLinReg model(inputSet, outputSet);
     // alg.printVector(model.modelSetTest(inputSet));
 
-    // // MULIVARIATE LINEAR REGRESSION
+    // MULIVARIATE LINEAR REGRESSION
     // std::vector<std::vector<double>> inputSet = {{1,2,3,4,5,6,7,8,9,10}, {3,5,9,12,15,18,21,24,27,30}};
     // std::vector<double> outputSet = {2,4,6,8,10,12,14,16,18,20};
     // LinReg model(alg.transpose(inputSet), outputSet); // Can use Lasso, Ridge, ElasticNet Reg
-    // model.normalEquation(); 
+    //model.normalEquation(); 
     // model.gradientDescent(0.001, 30000, 1);
     // model.SGD(0.001, 30000, 1);
     // model.MBGD(0.001, 10000, 2, 1);
     // alg.printVector(model.modelSetTest((alg.transpose(inputSet))));
     // std::cout << "ACCURACY: " << 100 * model.score() << "%" << std::endl;
 
-    // // LOGISTIC REGRESSION
-    // std::vector<std::vector<double>> inputSet; 
-    // std::vector<double> outputSet; 
-    // data.setData(30, "/Users/marcmelikyan/Desktop/Data/BreastCancer.csv", inputSet, outputSet);
-    // LogReg model(inputSet, outputSet); 
-    // //model.SGD(0.1, 50000, 0);
+    // LOGISTIC REGRESSION
+    std::vector<std::vector<double>> inputSet; 
+    std::vector<double> outputSet; 
+    data.setData(30, "/Users/marcmelikyan/Desktop/Data/BreastCancer.csv", inputSet, outputSet);
+    LogReg model(inputSet, outputSet); 
+    model.SGD(0.001, 100000, 0);
     // model.MLE(0.1, 10000, 0);
-    // alg.printVector(model.modelSetTest(inputSet));
-    // std::cout << "ACCURACY: " << 100 * model.score() << "%" << std::endl;
+    alg.printVector(model.modelSetTest(inputSet));
+    std::cout << "ACCURACY: " << 100 * model.score() << "%" << std::endl;
 
     // // PROBIT REGRESSION
     // std::vector<std::vector<double>> inputSet;