Lab 2: Fixed error in assignment 3

lab2/assignment3.R

@@ -5,14 +5,14 @@ data = read.csv("communities.csv")
 
 # ----1.----
 
-columns_to_scale <- setdiff(names(data), "ViolentCrimesPerPop")
+feature_column_names <- setdiff(names(data), "ViolentCrimesPerPop")
 
-scaler <- preProcess(data[, columns_to_scale])
+scaler <- preProcess(data[, feature_column_names])
 
 data_scaled <- data
-data_scaled[, columns_to_scale] <- predict(scaler, data[, columns_to_scale])
+data_scaled[, feature_column_names] <- predict(scaler, data[, feature_column_names])
 
-X <- data_scaled[, columns_to_scale]
+X <- data_scaled[, feature_column_names]
 y <- data_scaled$ViolentCrimesPerPop
 
 covariance_matrix <- cov(X)
@@ -65,18 +65,17 @@ id = sample(1:n, floor(n * 0.5))
 train = data[id, ]
 test = data[-id, ]
 
-scaler3 <- preProcess(train[, columns_to_scale])
+scaler3 <- preProcess(train)
 
-train_scaled <- train
-train_scaled[, columns_to_scale] <- predict(scaler3, train[, columns_to_scale])
+train_scaled<- predict(scaler3, train)
 
 test_scaled <- test
-test_scaled[, columns_to_scale] <- predict(scaler3, test[, columns_to_scale])
+test_scaled <- predict(scaler3, test)
 
-X_train <- train_scaled[, columns_to_scale]
+X_train <- train_scaled[, feature_column_names]
 y_train <- train_scaled$ViolentCrimesPerPop
 
-X_test <- test_scaled[, columns_to_scale]
+X_test <- test_scaled[, feature_column_names]
 y_test <- test_scaled$ViolentCrimesPerPop
 
 model <- lm(ViolentCrimesPerPop ~ . - 1, data = train_scaled)
@@ -102,8 +101,8 @@ cost_function <- function(theta) {
   train_error <- MSE(y_train, as.matrix(X_train) %*% theta)
   test_error <- MSE(y_test, as.matrix(X_test) %*% theta)
   
-  .GlobalEnv$train_errors <- c(train_errors, train_error)
-  .GlobalEnv$test_errors <- c(test_errors, test_error)
+  train_errors <<- c(train_errors, train_error)
+  test_errors <<- c(test_errors, test_error)
   
   return(train_error)
 }
@@ -112,7 +111,7 @@ res <- optim(par = rep(0, ncol(X_train)),
              fn = cost_function,
              method = "BFGS",)
 
-range <- c(train_errors[500:length(train_errors)], test_errors[500:length(test_errors)])
+range <- c(train_errors[501:length(train_errors)], test_errors[501:length(test_errors)])
 
 plot(
   train_errors,
@@ -122,7 +121,8 @@ plot(
   xlab = "Iteration",
   main = "Training and Test Errors over Iterations",
   lty = 1,
-  ylim = c(min(range), max(range))
+  ylim = c(min(range), max(range)),
+  xlim = c(501,length(train_errors)),
 )
 
 points(test_errors,
@@ -137,7 +137,7 @@ legend(
   lty = c(1, 2)
 )
 
-early_stopping <- which.min(test_errors[500:length(test_errors)]) + 500
+early_stopping <- which.min(test_errors)
 early_stopping
 
 test_errors[early_stopping]
@@ -147,4 +147,4 @@ optimal_train_mse <- res$value
 optimal_train_mse
 
 optimal_test_mse <- MSE(y_test, as.matrix(X_test) %*% optimal_theta)
-optimal_test_mse
+optimal_test_mse
\ No newline at end of file