Added loss functions

src/machine_learning/loss_functions/losses.h

@@ -0,0 +1,96 @@
+#ifndef MEAN_SQUARED_ERROR_H
+#define MEAN_SQUARED_ERROR_H
+
+#ifdef __cplusplus
+#include <iostream>
+#include <algorithm>
+#include <vector>
+#include <cmath>
+#include <cassert>
+#include <numbers>
+#endif
+
+/**
+ * @brief losses namespace that contains a couple of useful losses in machine learning
+ *
+ */
+namespace losses {
+
+  /**
+   * @brief sigmoid activation function
+   * @param x: double, the input parameter
+   * @return double: the sigmoid output
+   */
+  double sigmoid(double x) {
+    return 1.0 / (1.0 + exp(-x));
+  }
+
+  /**
+   * @brief mean squared error function
+   * @param y: vector, the original labels
+   * @param y_hat: vector, the predicted labels
+   * @return double: the mean squared error 
+   */
+  double mean_squared_error(std::vector<double> const& y, std::vector<double> const& y_hat) {
+    assert(y.size() == y_hat.size());
+    size_t n = y.size();
+    double mse = 0.0;
+    for(size_t i = 0; i<n; i++) {
+        mse += std::powf(y[i] - y_hat[i], 2);
+    }
+    return mse / double(n);
+  }
+
+  /**
+   * @brief root mean squared error function
+   * @param y: vector, the original labels
+   * @param y_hat: vector, the predicted labels
+   * @return double: the root mean squared error
+   */
+  double root_mean_squared_error(std::vector<double> const& y, std::vector<double> const& y_hat) {
+    return std::sqrt(mean_squared_error(y, y_hat));
+  }
+
+   /**
+  * @brief mean absolute error function
+  * @param y: vector, the original labels
+  * @param y_hat: vector, the predicted labels
+  * @return double: the mean absolute error
+  */
+  double mean_absolute_error(std::vector<double> const& y, std::vector<double> const& y_hat) {
+    assert(y.size() == y_hat.size());
+    size_t n = y.size();
+    double mae = 0.0;
+    for(size_t i = 0; i<n; i++) {
+        mae += std::abs(y[i] - y_hat[i]);
+    }
+    return mae / double(n);
+  }
+
+  /**
+  * @brief binary crossentropy loss function
+  * @param y: vector, the original labels
+  * @param y_hat: vector, the predicted labels
+  * @return double: the binary crossentropy loss
+  */
+  double binary_crossentropy_loss(std::vector<double> const& y, std::vector<double> const& y_hat) {
+    assert(y.size() == y_hat.size());
+    for(auto & x : y) {
+        assert(x == 0.0 || x == 1.0);
+    }
+
+    size_t n = y.size();
+    double bce = 0.0, eps = 1e-15;
+    for(size_t i = 0; i<n; i++) {
+        double prob = sigmoid(y_hat[i]);
+        double clipped_y_hat = std::clamp(prob, eps, 1 - eps);
+        bce += (y[i]*log(clipped_y_hat) + (1-y[i])*log(1 - clipped_y_hat));
+    }
+    return -bce / double(n);
+  }
+
+}
+
+
+#endif
+

tests/machine_learning/loss_functions/losses.cc

@@ -0,0 +1,29 @@
+#define CATCH_CONFIG_MAIN
+#include "../../../src/machine_learning/loss_functions/losses.h"
+#include "../../../third_party/catch.hpp"
+
+using namespace losses;
+
+TEST_CASE("Testing mean squared error") {
+
+  std::vector<double> v1 { 1.23, 4.25, 4.4, 1.231, 5.567 };
+  std::vector<double> v2 { 4.56, 4.123, 1.234, 6.432, 5.555 };
+
+  REQUIRE(mean_squared_error(v1, v2) == Approx(9.6358263926).epsilon(1e-6));
+}
+
+TEST_CASE("Testing mean absolute error") {
+  std::vector<double> v1 { 1.23, 4.25, 4.4, 1.231, 5.567 };
+  std::vector<double> v2 { 4.56, 4.123, 1.234, 6.432, 5.555 };
+
+  REQUIRE(mean_absolute_error(v1, v2) == Approx(2.36720).epsilon(1e-6));
+}
+
+TEST_CASE("Testing binary crossentropy loss") {
+  std::vector<double> v1 { 0., 0., 0., 0., 1., 0., 0., 0., 0., 1. };
+  std::vector<double> v2 { 1.8055,  0.9193, -0.2527,  1.0489,  0.5396, -1.2046, -0.9479,  0.8274,
+        -0.0548, -0.1902 };
+
+  REQUIRE(binary_crossentropy_loss(v1, v2) == Approx(0.8834657559).epsilon(1e-6));
+}
+