10/14/24, 12:17 PM TA1_U1_DL.

ipynb - Colab

import numpy as np
import tensorflow as tf
from tensorflow.keras import layers, models
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler, LabelEncoder
from sklearn.datasets import load_iris
import matplotlib.pyplot as plt

# Load Iris dataset

iris = load_iris()
X = iris.data
y = iris.target

# Encode the target variable

encoder = LabelEncoder()
y = encoder.fit_transform(y)
y = tf.keras.utils.to_categorical(y, 3)

# Split dataset into training and testing sets

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

# Standardize the input data

scaler = StandardScaler()
X_train = scaler.fit_transform(X_train)
X_test = scaler.transform(X_test)

def build_model(activation_function='relu'):
model = models.Sequential()
model.add(layers.Dense(64, input_shape=(4,), activation=activation_function, kernel_initializer='he_normal')) # He Initialization
model.add(layers.BatchNormalization()) # Batch Normalization
model.add(layers.Dense(32, activation=activation_function))
model.add(layers.BatchNormalization()) # Batch Normalization
model.add(layers.Dense(3, activation='softmax')) # Output layer for classification

model.compile(optimizer='adam',
loss='categorical_crossentropy',
metrics=['accuracy'])
return model

# Experimenting with different batch sizes

batch_sizes = [8, 16, 32, 64]

history_dict = {}
for batch_size in batch_sizes:
print(f"Training with batch size: {batch_size}")
model = build_model(activation_function='relu')
history = model.fit(X_train, y_train,
epochs=10,
batch_size=batch_size,
validation_data=(X_test, y_test),
verbose=1)
history_dict[batch_size] = history

https://colab.research.google.com/drive/1ur9jOfNcwMOs-7Yproawg4jWxJFGv-xp#scrollTo=3H9BH2Brx50-&printMode=true 1/3
10/14/24, 12:17 PM TA1_U1_DL.ipynb - Colab
/ / p y _ y _
Epoch 5/10
4/4 ━━━━━━━━━━━━━━━━━━━━ 0s 12ms/step - accuracy: 0.7652 - loss: 0.5695 - val_accuracy: 0.6000 - val_loss: 0.7729
Epoch 6/10
4/4 ━━━━━━━━━━━━━━━━━━━━ 0s 13ms/step - accuracy: 0.8240 - loss: 0.4026 - val_accuracy: 0.6333 - val_loss: 0.6450
Epoch 7/10
4/4 ━━━━━━━━━━━━━━━━━━━━ 0s 12ms/step - accuracy: 0.8475 - loss: 0.3988 - val_accuracy: 0.7000 - val_loss: 0.5472
Epoch 8/10
4/4 ━━━━━━━━━━━━━━━━━━━━ 0s 11ms/step - accuracy: 0.8913 - loss: 0.2886 - val_accuracy: 0.9000 - val_loss: 0.4759
Epoch 9/10
4/4 ━━━━━━━━━━━━━━━━━━━━ 0s 12ms/step - accuracy: 0.9544 - loss: 0.2337 - val_accuracy: 0.9000 - val_loss: 0.4278
Epoch 10/10
4/4 ━━━━━━━━━━━━━━━━━━━━ 0s 11ms/step - accuracy: 0.9300 - loss: 0.2688 - val_accuracy: 0.9000 - val_loss: 0.3947
Training with batch size: 64
Epoch 1/10
2/2 ━━━━━━━━━━━━━━━━━━━━ 2s 242ms/step - accuracy: 0.2514 - loss: 1.3962 - val_accuracy: 0.4000 - val_loss: 1.0408
Epoch 2/10
2/2 ━━━━━━━━━━━━━━━━━━━━ 0s 36ms/step - accuracy: 0.3597 - loss: 1.1184 - val_accuracy: 0.5667 - val_loss: 0.9246
Epoch 3/10
2/2 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step - accuracy: 0.5809 - loss: 0.8581 - val_accuracy: 0.6333 - val_loss: 0.8295
Epoch 4/10
2/2 ━━━━━━━━━━━━━━━━━━━━ 0s 46ms/step - accuracy: 0.6368 - loss: 0.7838 - val_accuracy: 0.7000 - val_loss: 0.7524
Epoch 5/10
2/2 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step - accuracy: 0.7774 - loss: 0.5761 - val_accuracy: 0.8000 - val_loss: 0.6902
Epoch 6/10
2/2 ━━━━━━━━━━━━━━━━━━━━ 0s 35ms/step - accuracy: 0.8156 - loss: 0.4862 - val_accuracy: 0.8667 - val_loss: 0.6404
Epoch 7/10
2/2 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step - accuracy: 0.8701 - loss: 0.3987 - val_accuracy: 0.8667 - val_loss: 0.5990
Epoch 8/10
2/2 ━━━━━━━━━━━━━━━━━━━━ 0s 33ms/step - accuracy: 0.9132 - loss: 0.3316 - val_accuracy: 0.8667 - val_loss: 0.5647
Epoch 9/10
2/2 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step - accuracy: 0.9399 - loss: 0.2672 - val_accuracy: 0.8333 - val_loss: 0.5376
Epoch 10/10
2/2 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step - accuracy: 0.9458 - loss: 0.2328 - val_accuracy: 0.8333 - val_loss: 0.5160

# Plot training accuracy and validation accuracy for different batch sizes
plt.figure(figsize=(14, 7))

# Accuracy Plot
plt.subplot(1, 2, 1)
for batch_size in batch_sizes:
plt.plot(history_dict[batch_size].history['accuracy'], label=f'Batch Size {batch_size}')
plt.title('Training Accuracy vs Batch Size')
plt.xlabel('Epochs')
plt.ylabel('Accuracy')
plt.legend()

# Loss Plot
plt.subplot(1, 2, 2)
for batch_size in batch_sizes:
plt.plot(history_dict[batch_size].history['loss'], label=f'Batch Size {batch_size}')
plt.title('Training Loss vs Batch Size')
plt.xlabel('Epochs')
plt.ylabel('Loss')
plt.legend()

plt.show()

https://colab.research.google.com/drive/1ur9jOfNcwMOs-7Yproawg4jWxJFGv-xp#scrollTo=3H9BH2Brx50-&printMode=true 2/3
10/14/24, 12:17 PM TA1_U1_DL.ipynb - Colab

https://colab.research.google.com/drive/1ur9jOfNcwMOs-7Yproawg4jWxJFGv-xp#scrollTo=3H9BH2Brx50-&printMode=true 3/3