From f6a144987cc403de8ad6bddd055f756120f8d405 Mon Sep 17 00:00:00 2001
From: Ethan Werner <mail@ethanwerner.com>
Date: Wed, 17 Apr 2024 16:00:00 -0700
Subject: archival commit

---
 example/ann_xor.c | 48 ++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 48 insertions(+)
 create mode 100755 example/ann_xor.c

(limited to 'example/ann_xor.c')

diff --git a/example/ann_xor.c b/example/ann_xor.c
new file mode 100755
index 0000000..a39264c
--- /dev/null
+++ b/example/ann_xor.c
@@ -0,0 +1,48 @@
+// example_xor.c - Example XOR training
+
+#define ANN_IMPLEMENTATION
+#include "../include/ann.h"
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <time.h>
+
+int main(void) {
+  srand(time(0));
+
+  ann_t *ann = ann_init(3, (uint_t[]){2, 2, 1});
+  ann_set_activation(ann, SIGMOID, SIGMOID);
+  ann_random(ann);
+
+  fp_t input[4][2] = {{0, 0}, {0, 1}, {1, 0}, {1, 1}};
+  fp_t output[4] = {0, 0, 0, 0};
+  fp_t expected[4] = {0, 1, 1, 0};
+  fp_t error = 0;
+
+  for (int_t iteration = 0; iteration < 100000; iteration++) {
+    for (int_t training = 0; training < 4; training++) {
+      ann_propagation_forward(ann, (fp_t const *const)&input[training],
+                              &output[training]);
+
+      error += ann_error_total(&output[training], &expected[training],
+                               ann->layer_neuron_n[ann->layer_n - 1]);
+
+      // Train using backpropagation
+      ann_propagation_backward(ann, (fp_t const *const)&input[training],
+                               &output[training], &expected[training], 0.4);
+
+      // Train using numeric integration
+      // ann_train_numeric( ann, ( fp_t const * const ) &i[b], &t[b], 0.03 );
+    }
+  }
+
+  ann_print_weight(ann);
+
+  fputs("------------------------------------------------------------\n\n",
+        stderr);
+
+  for (int i = 0; i < 4; i++) {
+    ann_propagation_forward(ann, (fp_t const *const)&input[i], &output[i]);
+    ann_print_neuron(ann, (fp_t const *const)&input[i], &output[i]);
+  }
+}
-- 
cgit v1.3.1