refactor: re-structure folder

src/__init__.py

@@ -0,0 +1,4 @@
+"""
+Helper utilities and modules for Deep Learning experiments.
+"""
+

src/data_loader.py

@@ -0,0 +1,46 @@
+import numpy as np
+from tensorflow import keras
+from typing import Tuple
+
+
+def load_mnist_for_mlp() -> Tuple[Tuple[np.ndarray, np.ndarray], Tuple[np.ndarray, np.ndarray]]:
+    """
+    Memuat dataset MNIST dan melakukan normalisasi serta perataan (flatten)
+    menjadi 784 fitur untuk arsitektur MLP.
+
+    Returns:
+        (x_train, y_train), (x_test, y_test)
+    """
+    (x_train, y_train), (x_test, y_test) = keras.datasets.mnist.load_data()
+    
+    # Normalisasi nilai pixel dari 0-255 menjadi 0.0-1.0
+    x_train = x_train.astype("float32") / 255.0
+    x_test = x_test.astype("float32") / 255.0
+    
+    # Flatten 28x28 menjadi 784
+    x_train = x_train.reshape(-1, 784)
+    x_test = x_test.reshape(-1, 784)
+    
+    return (x_train, y_train), (x_test, y_test)
+
+
+def load_mnist_for_cnn() -> Tuple[Tuple[np.ndarray, np.ndarray], Tuple[np.ndarray, np.ndarray]]:
+    """
+    Memuat dataset MNIST dan melakukan normalisasi serta penambahan dimensi channel
+    (28, 28, 1) untuk arsitektur CNN.
+
+    Returns:
+        (x_train, y_train), (x_test, y_test)
+    """
+    (x_train, y_train), (x_test, y_test) = keras.datasets.mnist.load_data()
+    
+    # Normalisasi nilai pixel dari 0-255 menjadi 0.0-1.0
+    x_train = x_train.astype("float32") / 255.0
+    x_test = x_test.astype("float32") / 255.0
+    
+    # Reshape ke format (N, 28, 28, 1)
+    x_train = x_train.reshape(-1, 28, 28, 1)
+    x_test = x_test.reshape(-1, 28, 28, 1)
+    
+    return (x_train, y_train), (x_test, y_test)
+

src/models.py

@@ -0,0 +1,82 @@
+from typing import List
+from tensorflow import keras
+from tensorflow.keras import layers
+
+
+def create_mlp_model(hidden_layers: List[int], input_dim: int = 784, num_classes: int = 10) -> keras.Model:
+    """
+    Membangun model MLP sederhana sesuai daftar neuron per hidden layer.
+    """
+    model = keras.Sequential(name="mlp_model")
+    model.add(layers.Input(shape=(input_dim,)))
+
+    for neurons in hidden_layers:
+        model.add(layers.Dense(neurons, activation="relu"))
+
+    model.add(layers.Dense(num_classes, activation="softmax"))
+
+    model.compile(
+        optimizer="adam",
+        loss="sparse_categorical_crossentropy",
+        metrics=["accuracy"]
+    )
+    return model
+
+
+def create_regularized_mlp_model(
+    hidden_layers: List[int] = [256, 128, 64],
+    dropout_rate: float = 0.3,
+    input_dim: int = 784,
+    num_classes: int = 10
+) -> keras.Model:
+    """
+    Membangun model MLP dengan Batch Normalization dan Dropout.
+    """
+    model = keras.Sequential(name="mlp_regularized")
+    model.add(layers.Input(shape=(input_dim,)))
+
+    for neurons in hidden_layers:
+        model.add(layers.Dense(neurons))
+        model.add(layers.BatchNormalization())
+        model.add(layers.Activation("relu"))
+        model.add(layers.Dropout(dropout_rate))
+
+    model.add(layers.Dense(num_classes, activation="softmax"))
+
+    model.compile(
+        optimizer="adam",
+        loss="sparse_categorical_crossentropy",
+        metrics=["accuracy"]
+    )
+    return model
+
+
+def create_cnn_mnist_model(input_shape: tuple = (28, 28, 1), num_classes: int = 10) -> keras.Model:
+    """
+    Membangun model Convolutional Neural Network (CNN) untuk MNIST.
+    """
+    model = keras.Sequential(
+        name="cnn_mnist",
+        layers=[
+            layers.Input(shape=input_shape),
+            # Block 1
+            layers.Conv2D(32, (3, 3), activation="relu"),
+            layers.MaxPooling2D((2, 2)),
+            # Block 2
+            layers.Conv2D(64, (3, 3), activation="relu"),
+            layers.MaxPooling2D((2, 2)),
+            # Classifier
+            layers.Flatten(),
+            layers.Dense(64, activation="relu"),
+            layers.Dropout(0.3),
+            layers.Dense(num_classes, activation="softmax"),
+        ]
+    )
+
+    model.compile(
+        optimizer="adam",
+        loss="sparse_categorical_crossentropy",
+        metrics=["accuracy"]
+    )
+    return model
+

src/utils.py

@@ -0,0 +1,65 @@
+import os
+import time
+from typing import Dict, List
+import matplotlib.pyplot as plt
+from tensorflow import keras
+
+
+class EpochTimerCallback(keras.callbacks.Callback):
+    """
+    Keras callback untuk mencatat durasi waktu setiap epoch.
+    """
+    def __init__(self):
+        super().__init__()
+        self.epoch_times: List[float] = []
+        self._start: float = 0.0
+
+    def on_epoch_begin(self, epoch, logs=None):
+        self._start = time.time()
+
+    def on_epoch_end(self, epoch, logs=None):
+        self.epoch_times.append(time.time() - self._start)
+
+
+def ensure_dir(dir_path: str):
+    """Membuat direktori jika belum ada."""
+    os.makedirs(dir_path, exist_ok=True)
+
+
+def plot_training_history(
+    history,
+    title: str = "Training History",
+    save_path: str = None
+):
+    """
+    Membuat plot loss dan akurasi dari training history Keras.
+    """
+    fig, axes = plt.subplots(1, 2, figsize=(12, 4.5))
+
+    # Loss plot
+    axes[0].plot(history.history.get("loss", []), label="Train Loss")
+    if "val_loss" in history.history:
+        axes[0].plot(history.history["val_loss"], label="Val Loss", linestyle="--")
+    axes[0].set_title(f"{title} - Loss")
+    axes[0].set_xlabel("Epoch")
+    axes[0].set_ylabel("Loss")
+    axes[0].legend()
+    axes[0].grid(True, alpha=0.3)
+
+    # Accuracy plot
+    axes[1].plot(history.history.get("accuracy", []), label="Train Accuracy")
+    if "val_accuracy" in history.history:
+        axes[1].plot(history.history["val_accuracy"], label="Val Accuracy", linestyle="--")
+    axes[1].set_title(f"{title} - Accuracy")
+    axes[1].set_xlabel("Epoch")
+    axes[1].set_ylabel("Accuracy")
+    axes[1].legend()
+    axes[1].grid(True, alpha=0.3)
+
+    plt.tight_layout()
+    if save_path:
+        ensure_dir(os.path.dirname(save_path))
+        plt.savefig(save_path, dpi=130)
+        print(f"[INFO] Plot disimpan di: {save_path}")
+    plt.close()
+