11/27/24, 1:01 AM _DL_Lab5..
ipynb - Colab
Rushikesh Jadhav
232010013
DL Lab 5
Aim : Implementation of Regularization Techniques
import tensorflow as tf
from tensorflow.keras import layers, models
from sklearn.model_selection import train_test_split
from sklearn.datasets import fetch_california_housing
# Load California housing dataset
data = fetch_california_housing()
X, y = data.data, data.target
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=4
1. Dropout
add Code add Text
# Build model with Dropout
model_dropout = models.Sequential([
layers.Dense(64, activation='relu', input_shape=(X_train.shape[1],)),
layers.Dropout(0.5),
layers.Dense(64, activation='relu'),
layers.Dense(1)
])
model_dropout.compile(optimizer='adam', loss='mean_squared_error')
model_dropout.fit(X_train, y_train, epochs=10, validation_split=0.2)
Epoch 1/10
/usr/local/lib/python3.10/dist-packages/keras/src/layers/core/dense.py:87: UserWa
super().__init__(activity_regularizer=activity_regularizer, **kwargs)
413/413 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - loss: 15149.9199 - val_loss: 22.0540
Epoch 2/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 1051.6801 - val_loss: 42.0648
Epoch 3/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 574.1198 - val_loss: 13.8391
Epoch 4/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 236.6241 - val_loss: 43.7359
Epoch 5/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 2s 4ms/step - loss: 233.4340 - val_loss: 3.3071
Epoch 6/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 2s 4ms/step - loss: 91.3531 - val_loss: 39.9956
Epoch 7/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 4s 7ms/step - loss: 132.5156 - val_loss: 7.4772
Epoch 8/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 2s 5ms/step - loss: 47.8788 - val_loss: 8.1829
Epoch 9/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 48.0613 - val_loss: 15.4679
Epoch 10/10
https://colab.research.google.com/drive/1PuT4Pqwhlz8ys8ALxVP1rUKzyzVp0Mi3?usp=sharing… 1/4
11/27/24, 1:01 AM _DL_Lab5..ipynb - Colab
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 34.5671 - val_loss: 2.9151
<keras.src.callbacks.history.History at 0x7bcedf055540>
2. Early Stopping
# Early stopping callback
early_stopping = tf.keras.callbacks.EarlyStopping(monitor='val_loss', patience=3, rest
# Train the model with early stopping
model_dropout.fit(X_train, y_train, epochs=10, validation_split=0.2, callbacks=[early_
Epoch 1/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 22.5465 - val_loss: 3.3754
Epoch 2/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 10.0820 - val_loss: 4.1327
Epoch 3/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 6.0967 - val_loss: 2.1741
Epoch 4/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 4.9395 - val_loss: 2.2975
Epoch 5/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 3.6868 - val_loss: 1.5492
Epoch 6/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 3.2472 - val_loss: 1.4901
Epoch 7/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 2.9435 - val_loss: 1.5168
Epoch 8/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 2.2726 - val_loss: 1.3492
Epoch 9/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 2s 3ms/step - loss: 2.6524 - val_loss: 1.3322
Epoch 10/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step - loss: 2.0697 - val_loss: 1.1929
<keras.src.callbacks.history.History at 0x7bcede6444f0>
3. L1 regularization
# Build model with L2 regularization
model_l2 = models.Sequential([
layers.Dense(64, activation='relu', kernel_regularizer=tf.keras.regularizers.l2(0
layers.Dense(64, activation='relu', kernel_regularizer=tf.keras.regularizers.l2(0
layers.Dense(1)
])
model_l2.compile(optimizer='adam', loss='mean_squared_error')
model_l2.fit(X_train, y_train, epochs=10, validation_split=0.2)
Epoch 1/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - loss: 298.8849 - val_loss: 1.6367
Epoch 2/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 1.6259 - val_loss: 1.3696
Epoch 3/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 10.6650 - val_loss: 2.8881
https://colab.research.google.com/drive/1PuT4Pqwhlz8ys8ALxVP1rUKzyzVp0Mi3?usp=sharing… 2/4
11/27/24, 1:01 AM _DL_Lab5..ipynb - Colab
Epoch 4/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 2.9807 - val_loss: 1.6852
Epoch 5/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 5.3804 - val_loss: 1.1550
Epoch 6/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 1.3497 - val_loss: 1.3505
Epoch 7/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 20.1218 - val_loss: 5.1635
Epoch 8/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 1.6839 - val_loss: 1.3582
Epoch 9/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 2s 3ms/step - loss: 1.1762 - val_loss: 1.2227
Epoch 10/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step - loss: 1.0459 - val_loss: 1.3154
<keras.src.callbacks.history.History at 0x7bcede6dc2e0>
4. L2 regularization
# Build model with L2 regularization
model_l2 = models.Sequential([
layers.Dense(64, activation='relu', kernel_regularizer=tf.keras.regularizers.l2(0
layers.Dense(64, activation='relu', kernel_regularizer=tf.keras.regularizers.l2(0
layers.Dense(1)
])
model_l2.compile(optimizer='adam', loss='mean_squared_error')
model_l2.fit(X_train, y_train, epochs=10, validation_split=0.2)
Epoch 1/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - loss: 221.3601 - val_loss: 2.4019
Epoch 2/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 144.9611 - val_loss: 1.4573
Epoch 3/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 1.9867 - val_loss: 1.2558
Epoch 4/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 1.5741 - val_loss: 10.6253
Epoch 5/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 1.8362 - val_loss: 98.7645
Epoch 6/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 156.8823 - val_loss: 1.2026
Epoch 7/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 3.0938 - val_loss: 1.1754
Epoch 8/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 3.0639 - val_loss: 1.9080
Epoch 9/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 2s 3ms/step - loss: 1.5629 - val_loss: 2.1380
Epoch 10/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step - loss: 15.6031 - val_loss: 7.3652
<keras.src.callbacks.history.History at 0x7bcee04c3880>
5. Elastic Net
https://colab.research.google.com/drive/1PuT4Pqwhlz8ys8ALxVP1rUKzyzVp0Mi3?usp=sharing… 3/4
11/27/24, 1:01 AM _DL_Lab5..ipynb - Colab
# Build model with Elastic Net (L1 and L2 regularization)
model_elastic_net = models.Sequential([
layers.Dense(64, activation='relu', kernel_regularizer=tf.keras.regularizers.l1_l2
layers.Dense(64, activation='relu', kernel_regularizer=tf.keras.regularizers.l1_l2
layers.Dense(1)
])
model_elastic_net.compile(optimizer='adam', loss='mean_squared_error')
model_elastic_net.fit(X_train, y_train, epochs=10, validation_split=0.2)
Epoch 1/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - loss: 4133.4990 - val_loss: 5.0145
Epoch 2/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 5.2582 - val_loss: 5.3149
Epoch 3/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 2s 3ms/step - loss: 4.3300 - val_loss: 3.9747
Epoch 4/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step - loss: 7.3997 - val_loss: 68.9548
Epoch 5/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - loss: 11.6055 - val_loss: 3.8571
Epoch 6/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 101.9356 - val_loss: 5.4602
Epoch 7/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 10.6571 - val_loss: 3.8880
Epoch 8/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 5.3400 - val_loss: 3.7929
Epoch 9/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 16.1976 - val_loss: 3.6659
Epoch 10/10
413/413 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step - loss: 5.7418 - val_loss: 3.6570
<keras.src.callbacks.history.History at 0x7bced6253910>
https://colab.research.google.com/drive/1PuT4Pqwhlz8ys8ALxVP1rUKzyzVp0Mi3?usp=sharing… 4/4
