From 777ac49980dbbad3769b1ee0db67420bd1074081 Mon Sep 17 00:00:00 2001
From: sosokker <sirin_29@hotmail.co.th>
Date: Wed, 23 Aug 2023 23:50:19 +0700
Subject: [PATCH] Add train/predict method

---
 src/utils/utils.cpp | 48 +++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 48 insertions(+)
 create mode 100644 src/utils/utils.cpp

diff --git a/src/utils/utils.cpp b/src/utils/utils.cpp
new file mode 100644
index 0000000..c853a7a
--- /dev/null
+++ b/src/utils/utils.cpp
@@ -0,0 +1,48 @@
+#include <iostream>
+#include <vector>
+#include <functional>
+#include <Eigen/Dense>
+#include <Layers/Layers.hpp>
+
+Eigen::VectorXd predict(const std::vector<Layer*>& network, const Eigen::VectorXd& input) {
+    Eigen::VectorXd output = input;
+    for (const auto& layer : network) {
+        layer->forward(output);
+        output = layer->getOutput();
+    }
+    return output;
+}
+
+void train(std::vector<Layer*>& network,
+           const std::function<Eigen::VectorXd(const Eigen::VectorXd&, const Eigen::VectorXd&)>& loss,
+           const std::function<Eigen::VectorXd(const Eigen::VectorXd&, const Eigen::VectorXd&)>& loss_prime,
+           const std::vector<Eigen::VectorXd>& x_train,
+           const std::vector<Eigen::VectorXd>& y_train,
+           int epochs = 1000,
+           double learning_rate = 0.01,
+           bool verbose = true) {
+
+    for (int e = 0; e < epochs; ++e) {
+        double total_error = 0.0;
+        for (size_t i = 0; i < x_train.size(); ++i) {
+            // Forward pass
+            Eigen::VectorXd output = predict(network, x_train[i]);
+
+            // Compute loss
+            Eigen::VectorXd error = loss(y_train[i], output);
+            total_error += error.sum();
+
+            // Backward pass
+            Eigen::VectorXd grad = loss_prime(y_train[i], output);
+            for (auto it = network.rbegin(); it != network.rend(); ++it) {
+                (*it)->backward(grad, learning_rate);
+                grad = (*it)->getInputGradient();
+            }
+        }
+
+        double average_error = total_error / x_train.size();
+        if (verbose) {
+            std::cout << e + 1 << "/" << epochs << ", error=" << average_error << std::endl;
+        }
+    }
+}
\ No newline at end of file