New functions and updated README.md 🚀

* R-squared (rsq) with the possibility of adjusting for parameters
	- Unit tests have been made.

* Huber loss (huberloss)
	- Unit tests have been made.

* README.md have been rewritten and the layout have been updated.
	- It needs a logo too.

NOTE: All the functions where inputs are not altered could
benefit from adding a const declaration. This will happen
at a later point.

R/RcppExports.R

@@ -1,6 +1,19 @@
 # Generated by using Rcpp::compileAttributes() -> do not edit by hand
 # Generator token: 10BE3573-1514-4C36-9D1C-5A225CD40393
 
+#' Huber Loss
+#'
+#' Calculate the huber loss of two <[numeric]> vectors.
+#'
+#' @param actual A <[numeric]>-vector of length N.
+#' @param predicted A <[numeric]>-vector of length N.
+#' @param delta A <[numeric]>-vector of length 1. 1 by default.
+#'
+#' @returns A <[numeric]>-value of length 1.
+huberloss <- function(actual, predicted, delta = 1) {
+    .Call(`_SLmetrics_huberloss`, actual, predicted, delta)
+}
+
 #' Mean Absolute Error (MAE)
 #'
 #' Calculate the MAE of two <[numeric]>-vectors
@@ -57,3 +70,17 @@ rmsle <- function(actual, predicted) {
     .Call(`_SLmetrics_rmsle`, actual, predicted)
 }
 
+#' R squared
+#'
+#' Calculate the R squared of two <[numeric]> vectors.
+#'
+#' @param actual A <[numeric]>-vector of length N.
+#' @param predicted A <[numeric]>-vector of length N.
+#' @param k A <[numeric]>-vector of length 1. 0 by default. If k>0
+#' the function returns the adjusted R squared.
+#'
+#' @returns A <[numeric]>-value of length 1.
+rsq <- function(actual, predicted, k = 0) {
+    .Call(`_SLmetrics_rsq`, actual, predicted, k)
+}
+

README.Rmd

@@ -20,10 +20,9 @@ knitr::opts_chunk$set(
 [![Codecov test coverage](https://codecov.io/gh/serkor1/MLmetrics/graph/badge.svg)](https://app.codecov.io/gh/serkor1/MLmetrics)
 <!-- badges: end -->
 
-{SLmetrics} is a collection of performance evaluation metrics for regression and classification models written in c++ and [{Rcpp}](https://github.com/RcppCore/Rcpp)
-
-## Example: Linear Regression
+{SLmetrics} is a collection of (lightning fast) performance evaluation metrics for regression and classification models written in c++ and [{Rcpp}](https://github.com/RcppCore/Rcpp). It's like using a supercharged yardstick to measure model performance, without the risk of soft to super-hard deprecations.
 
+## :information_source: Basic usage 
 ```{r}
 # 0) load {SLmetrics}
 library(SLmetrics)
@@ -36,13 +35,11 @@ model <- lm(
 
 # 2) evaluate RMSE
 rmse(
-  actual    = fitted(model),
-  predicted = predict(model) 
+  actual    = mtcars$mpg,
+  predicted = fitted(model)
 )
 ```
-
-
-
+</details>
 
 ## :information_source: Installation
 

README.md

@@ -10,11 +10,13 @@
 coverage](https://codecov.io/gh/serkor1/MLmetrics/graph/badge.svg)](https://app.codecov.io/gh/serkor1/MLmetrics)
 <!-- badges: end -->
 
-{SLmetrics} is a collection of performance evaluation metrics for
-regression and classification models written in c++ and
-[{Rcpp}](https://github.com/RcppCore/Rcpp)
+{SLmetrics} is a collection of (lightning fast) performance evaluation
+metrics for regression and classification models written in c++ and
+[{Rcpp}](https://github.com/RcppCore/Rcpp). It’s like using a
+supercharged yardstick to measure model performance, without the risk of
+soft to super-hard deprecations.
 
-## Example: Linear Regression
+## :information_source: Basic usage
 
 ``` r
 # 0) load {SLmetrics}
@@ -28,12 +30,14 @@ model <- lm(
 
 # 2) evaluate RMSE
 rmse(
-  actual    = fitted(model),
-  predicted = predict(model) 
+  actual    = mtcars$mpg,
+  predicted = fitted(model)
 )
-#> [1] 1.385601e-14
+#> [1] 2.146905
 ```
 
+</details>
+
 ## :information_source: Installation
 
 ### :shield: Stable version

src/RcppExports.cpp

@@ -10,6 +10,19 @@ Rcpp::Rostream<true>&  Rcpp::Rcout = Rcpp::Rcpp_cout_get();
 Rcpp::Rostream<false>& Rcpp::Rcerr = Rcpp::Rcpp_cerr_get();
 #endif
 
+// huberloss
+double huberloss(const NumericVector& actual, const NumericVector& predicted, const double delta);
+RcppExport SEXP _SLmetrics_huberloss(SEXP actualSEXP, SEXP predictedSEXP, SEXP deltaSEXP) {
+BEGIN_RCPP
+    Rcpp::RObject rcpp_result_gen;
+    Rcpp::RNGScope rcpp_rngScope_gen;
+    Rcpp::traits::input_parameter< const NumericVector& >::type actual(actualSEXP);
+    Rcpp::traits::input_parameter< const NumericVector& >::type predicted(predictedSEXP);
+    Rcpp::traits::input_parameter< const double >::type delta(deltaSEXP);
+    rcpp_result_gen = Rcpp::wrap(huberloss(actual, predicted, delta));
+    return rcpp_result_gen;
+END_RCPP
+}
 // mae
 double mae(const Rcpp::NumericVector& actual, const Rcpp::NumericVector& predicted);
 RcppExport SEXP _SLmetrics_mae(SEXP actualSEXP, SEXP predictedSEXP) {
@@ -58,12 +71,27 @@ BEGIN_RCPP
     return rcpp_result_gen;
 END_RCPP
 }
+// rsq
+double rsq(const NumericVector& actual, const NumericVector& predicted, const double k);
+RcppExport SEXP _SLmetrics_rsq(SEXP actualSEXP, SEXP predictedSEXP, SEXP kSEXP) {
+BEGIN_RCPP
+    Rcpp::RObject rcpp_result_gen;
+    Rcpp::RNGScope rcpp_rngScope_gen;
+    Rcpp::traits::input_parameter< const NumericVector& >::type actual(actualSEXP);
+    Rcpp::traits::input_parameter< const NumericVector& >::type predicted(predictedSEXP);
+    Rcpp::traits::input_parameter< const double >::type k(kSEXP);
+    rcpp_result_gen = Rcpp::wrap(rsq(actual, predicted, k));
+    return rcpp_result_gen;
+END_RCPP
+}
 
 static const R_CallMethodDef CallEntries[] = {
+    {"_SLmetrics_huberloss", (DL_FUNC) &_SLmetrics_huberloss, 3},
     {"_SLmetrics_mae", (DL_FUNC) &_SLmetrics_mae, 2},
     {"_SLmetrics_mse", (DL_FUNC) &_SLmetrics_mse, 2},
     {"_SLmetrics_rmse", (DL_FUNC) &_SLmetrics_rmse, 2},
     {"_SLmetrics_rmsle", (DL_FUNC) &_SLmetrics_rmsle, 2},
+    {"_SLmetrics_rsq", (DL_FUNC) &_SLmetrics_rsq, 3},
     {NULL, NULL, 0}
 };
 

src/regression_HUBER.cpp

@@ -0,0 +1,39 @@
+#include <Rcpp.h>
+using namespace Rcpp;
+
+//' Huber Loss
+//'
+//' Calculate the huber loss of two <[numeric]> vectors.
+//'
+//' @param actual A <[numeric]>-vector of length N.
+//' @param predicted A <[numeric]>-vector of length N.
+//' @param delta A <[numeric]>-vector of length 1. 1 by default.
+//'
+//' @returns A <[numeric]>-value of length 1.
+// [[Rcpp::export]]
+double huberloss(
+    const NumericVector& actual,
+    const NumericVector& predicted,
+    const double delta = 1) {
+
+  // Get the size of the vectors
+  const std::size_t n = actual.size();
+
+  // Initialize the output variable
+  double output = 0.0;
+
+  // Iterate over the vectors and compute the Huber loss
+  for (std::size_t i = 0; i < n; ++i) {
+    double diff = actual[i] - predicted[i];
+    double abs_diff = std::abs(diff);
+
+    if (abs_diff <= delta) {
+      output += 0.5 * diff * diff;
+    } else {
+      output += delta * (abs_diff - 0.5 * delta);
+    }
+  }
+
+  // Return the mean Huber loss
+  return output / n;
+}

src/regression_RSQ.cpp

@@ -0,0 +1,40 @@
+#include <Rcpp.h>
+using namespace Rcpp;
+
+//' R squared
+//'
+//' Calculate the R squared of two <[numeric]> vectors.
+//'
+//' @param actual A <[numeric]>-vector of length N.
+//' @param predicted A <[numeric]>-vector of length N.
+//' @param k A <[numeric]>-vector of length 1. 0 by default. If k>0
+//' the function returns the adjusted R squared.
+//'
+//' @returns A <[numeric]>-value of length 1.
+// [[Rcpp::export]]
+double rsq(
+    const NumericVector& actual,
+    const NumericVector& predicted,
+    const double k = 0) {
+
+  // Get the size of the vectors
+  const std::size_t n = actual.size();
+
+  // Calculate the mean of actual values
+  double mean_actual = std::accumulate(actual.begin(), actual.end(), 0.0) / n;
+
+  // Calculate SSE and SSR
+  double SSE = 0.0;
+  double SST = 0.0;
+
+  for (std::size_t i = 0; i < n; ++i) {
+    const double actual_val = actual[i];
+    const double predicted_val = predicted[i];
+
+    SSE += (actual_val - predicted_val) * (actual_val - predicted_val);
+    SST += (actual_val - mean_actual) * (actual_val - mean_actual);
+  }
+
+  // Calculate R-squared
+  return 1.0 - (SSE / SST) * ((n - 1))/(n - (k + 1));
+}

tests/testthat/test-huber.R

@@ -0,0 +1,43 @@
+# script: Huber Loss test
+# author: Serkan Korkmaz, [email protected]
+# date: 2024-08-26
+# objective: Test that the huber loss function
+# returns a non-zero value
+# script start;
+
+
+testthat::test_that(
+  desc = "`huberloss`-function returns non-zero postive values",
+  code = {
+
+    # 0) generate values
+    # from a normal distribution
+    actual <-  rnorm(
+      n = 1e2
+    )
+
+    predicted <- actual + rnorm(
+      n = 1e2
+    )
+
+
+    # 1) calculate the
+    # rsq using rsq()-function
+    output <- testthat::expect_no_condition(
+      huberloss(
+        predicted,
+        actual,
+        delta = 5
+      )
+    )
+
+    # 2) test that the value
+    # is greater than 0
+    testthat::expect_true(
+      output > 0
+    )
+
+  }
+)
+
+# script end;

tests/testthat/test-rsq.R

@@ -0,0 +1,41 @@
+# script: RSQ Tests
+# author: Serkan Korkmaz, [email protected]
+# date: 2024-08-26
+# objective: Test that rsq returns a non-zero
+# positive value for all values passed into it
+# script start;
+
+testthat::test_that(
+  desc = "`rsq`-function returns non-zero postive values",
+  code = {
+
+    # 0) generate values
+    # from a normal distribution
+    actual <-  rnorm(
+      n = 1e2
+    )
+
+    predicted <- actual + rnorm(
+      n = 1e2
+    )
+
+
+    # 1) calculate the
+    # rsq using rsq()-function
+    output <- testthat::expect_no_condition(
+      rsq(
+        predicted,
+        actual
+      )
+    )
+
+    # 2) test that the value
+    # is greater than 0
+    testthat::expect_true(
+      output > 0
+    )
+
+  }
+)
+
+# script end;