feature(main): working on the design. i dont know how machine learning works pfff...main

4 files changed, 154 insertions, 31 deletions

diff --git a/Makefile b/Makefile
index 2f5d478..9846889 100644
--- a/Makefile
+++ b/Makefile
@@ -1,27 +1,42 @@
-BIN = build/tb_ml
-SRC = source/tb_ml/tb_ml.c
+# Build configuration
+BIN := build/tb_ml
+SRC := source/tb_ml/tb_ml.c
+OBJ := build/tb_ml.o
+DEP := build/tb_ml.d
 
-# CC = gcc
-CC = clang
+# Compiler flags
+CC ?= clang
+CFLAGS := -Wall -Wextra -Wpedantic -Wno-unused-function -g --std=c99 -fno-omit-frame-pointer
+LDFLAGS := -lm
 
-COMPILER := $(shell $(CC) --version | grep -o "gcc\|clang" | head -1)
+# Optimization level
+OPT ?= 0
+CFLAGS += -O$(OPT)
 
-# check for compile optimizations per compiler
-ifeq ($(COMPILER),gcc)
-    CFLAGS = -Wall -Wextra -Wpedantic -Wno-unused-function -g -Werror -O0
-else ifeq ($(COMPILER),clang)
-    CFLAGS = -Wall -Wextra -Wpedantic -Wno-unused-function -g -O0
-else
-    CFLAGS = -Wall -Wextra -Wpedantic -Wno-unused-function -g -O0
-endif
+# Build directory
+$(shell mkdir -p build)
 
-$(BIN): $(SRC)
-        mkdir -p build
-        $(CC) $(CFLAGS) $< -o $@
+# Targets
+.PHONY: all run clean help
+
+all: $(BIN)
+
+$(BIN): $(OBJ)
+        $(CC) $(CFLAGS) $(LDFLAGS) $< -o $@
+
+$(OBJ): $(SRC)
+        $(CC) $(CFLAGS) -MT $@ -MMD -MP -MF $(DEP) -c $< -o $@
+
+-include $(DEP)
 
 run: $(BIN)
-        $(BIN)
+        ./$(BIN)
 
-.PHONY: clean
 clean:
         rm -rf build
+
+help:
+        @echo "make [OPT=0|1|2|3]"
+        @echo "  all    - Build (default)"
+        @echo "  run    - Build and run"
+        @echo "  clean  - Remove artifacts"
diff --git a/README.md b/README.md
index b059617..c4a4964 100644
--- a/README.md
+++ b/README.md
@@ -1,7 +1,30 @@
-TB_ML
+# TB_ML
 
-Attempt at writing a machine learning library in c using tb_db for the input data.
+An attempt at writing a machine learning library in C, using `tb_db` as the data input layer.
+[tb_db](https://git.nsrddyn.com/tb_db.git/)
 
-pseudo random number generation
-https://www.youtube.com/watch?v=OlSYfj8VZi0
-https://www.pcg-random.org/
+---
+
+## References
+
+- [Computerphile – Random Numbers (YouTube)](https://www.youtube.com/watch?v=OlSYfj8VZi0)
+- [PCG Random – O'Neill (pcg-random.org)](https://www.pcg-random.org/)
+- [Pattern Recognition and Machine Learning – Bishop (PDF)](https://www.cs.uoi.gr/~arly/courses/ml/tmp/Bishop_book.pdf)
+- [Mean Squared Error – Wikipedia](https://en.wikipedia.org/wiki/Mean_squared_error)
+- [Lineaire Benadering – Wikipedia (NL)](https://nl.wikipedia.org/wiki/Lineaire_benadering)
+- [Partial Differential Equations – Wikipedia](https://en.wikipedia.org/wiki/Partial_differential_equation)
+
+- [ChatGPT]  used for navigating unfamiliar topics and finding relevant material
+- [Claude]  used for mathematical implementations because I suck
+
+---
+
+## Dependencies
+
+- [`tb_db`](https://git.nsrddyn.com/tb_db) — custom CSV query engine / database layer
+
+---
+
+## Status
+
+Work in progress.
diff --git a/source/tb_ml/tb_mh.h b/source/tb_ml/tb_mh.h
new file mode 100644
index 0000000..91e1e7b
--- /dev/null
+++ b/source/tb_ml/tb_mh.h
@@ -0,0 +1,43 @@
+#ifndef TB_ML_H
+#define TB_ML_H
+
+
+//- supervised learning model by abdellah el morabit
+//- from <math.h>
+#define M_E  2.7182818284590452354      /* e */
+
+typedef struct training_sample trainig_sample;
+struct training_sample
+{
+
+};
+
+typedef struct prediction_data predicition_data;
+struct prediction_data
+{
+    f32     w;
+    f32     y;
+};
+
+
+typedef struct prediction_category prediction_category;
+struct category
+{
+    string8 name;
+    predicition_data data;
+};
+
+f(f32 d) {
+    return d * d;
+}
+
+internal f32
+derivative(f32 x, f32 (*f)(f32)) {
+
+  f32 h = 1e-6; // small interval to differnetiate the scope
+  return (f(x + h) - f(x - h)) / (2.0 * h);
+}
+
+
+
+#endif
diff --git a/source/tb_ml/tb_ml.c b/source/tb_ml/tb_ml.c
index cf9625a..8211a60 100644
--- a/source/tb_ml/tb_ml.c
+++ b/source/tb_ml/tb_ml.c
@@ -1,20 +1,62 @@
-#define BASE_UNITY
+#define BASE_IMPLEMENTATION
 #include "../base/base_include.h"
 
-#if 0
+#ifdef BTREE_IMPLEMENTATION
+internal void
+btree_collect(btree_node *node, (void *) *keys, (void *) array *) {
 
-#include "../third_party/btree_impl.h"
+    if (!n) return;
+
+    for (s32 index = 0; index < node->count; ++index)
+    {
+        TODO(nasr): traverse the tree to use the tree as training data
+    }
+}
 #endif
 
 
+/**
+* calculating a derviate requires a building an algebraic system
+* so we do a dump approximation of what the derivative could be
+* we want this to be able to normalize the data which is necessary for proper training
+**/
+#if 1
+internal f32
+internal training_sample_data
+normalize_training_data_btree(btree *btree) {
+
+
+
+}
+
+#endif
 
-//- dataset is an implementation of the btree
 
-int main(int c, char **v)
-{
-    unused(c);
-    unused(v);
+// using gradient descent
+internal void
+train(f32 *weights, b_tree, s32 learning_rate) {
+
+    btree_collect(btree_node *node, (void *) *keys, (void *) array *);
+
+    for (s32 epoch = 0; epoch < learning_rate; ++epoch)
+    {
+
+    }
+
+}
+
+internal f32
+predict(f32 b1, f32 b2) {
+    return b1 + (b2 * log(b1)) + M_E;
+}
+
+//- dataset is an implementation of the btree
+int main(void) {
 
+    for (s32 index = 0; index < 10; ++index)
+    {
+        printf("value [%4.f]\n",predict(10, 10));
+    }
 
     return 0;
 }