lOMoARcPSD|32921733

DEEP Learning R18 Jntuh LAB Manual

Deep Leaning (Jawaharlal Nehru Technological University, Hyderabad)

Studocu is not sponsored or endorsed by any college or university

Downloaded by sravanthi JNTUM ([email protected])
 lOMoARcPSD|32921733

R18 B.Tech. AI&ML DEEP LEARNING LAB JNTU Hyderabad

1. Setting up the Spyder IDE Environment and Executing a Python

Program
AIM: Setting up the Spyder IDE Environment and Executing a Python
Program
DESCRIPTION:
To set up the Spyder IDE (Integrated Development Environment) and
execute a Python program, follow these steps:
1. Install Download and install Python from official website:
Python https://www.python.org/downloads/
2. Install Anaconda, a popular Python distribution for data science,
Spyder includes Spyder and pre-installed packages for scientific
computing and analysis. Download from the official website:
https://www.anaconda.com/products/individual and follow
installation instructions.
3. Launch Launch Spyder IDE after installing Anaconda.
Spyder
4. Create In Spyder, you can create a new Python file or open an existing
or open one. To create a new file, go to "File" > "New File" or use the
a keyboard shortcut Ctrl+N (Command+N on Mac).
Python Alternatively, you can open an existing Python file by going to
file "File" > "Open" or using the keyboard shortcut Ctrl+O
(Command+O on Mac).
5. Write Write Python code in editor window for simple program. For
your example,
Python print("Hello, World!")
code
6. Save the Save your Python file with the ".py" extension, for example,
Python "hello.py". You can do this by going to "File" > "Save" or
file using the keyboard shortcut Ctrl+S (Command+S on Mac).
7. Execute To execute your Python program, go to "Run" > "Run" or press
the the F5 key. The output of your program will appear in the
Python console pane at the bottom of the Spyder IDE.
progra
m

That's it! You have set up the Spyder IDE environment and executed a
Python program. You can now continue writing and executing more complex
Python code within Spyder.

Downloaded by sravanthi JNTUM ([email protected])

 lOMoARcPSD|32921733

R18 B.Tech. AI&ML DEEP LEARNING LAB JNTU Hyderabad

2. Installing Keras, Tensorflow and Pytorch libraries and making use of

them
AIM:
Installing Keras, Tensorflow and Pytorch libraries and making use of them
DESCRIPTION:
To install Keras, TensorFlow, and PyTorch libraries and make use of them, you
can follow the steps below:
1. Install Ensure Python is installed and download the latest
Python and version from the official website:
pip (if not https://www.python.org/downloads/
already Python comes pre-installed with pip; install if
installed) unavailable via pip website:
https://pip.pypa.io/en/stable/installing/
2. Install Open a command prompt or terminal and run the
TensorFlow following command to install TensorFlow using pip:
pip install tensorflow
3. Install Keras Keras integrated into TensorFlow, ensuring automatic
installation during installation. However, you can
explicitly install Keras using pip: pip install keras
4. Install Install PyTorch by visiting official website, selecting
PyTorch appropriate command based on system configuration:
https://pytorch.org/get-started/locally/
For example, to install the CPU-only version of PyTorch
using pip, you can run: pip install torch torchvision
5. Verify After installing the libraries, you can verify that
installations everything is set up correctly by launching a Python
interpreter or creating a Python script and importing the
libraries:
import tensorflow as tf
import keras
import torch
print("TensorFlow version:", tf.__version__)
print("Keras version:", keras.__version__)
print("PyTorch version:", torch.__version__)
This code will output the versions of the installed
libraries, confirming that everything is installed correctly.
6. Using the Install libraries, use for machine learning models
libraries training.
Here's a basic example of how you can create a simple
3

Downloaded by sravanthi JNTUM ([email protected])

 lOMoARcPSD|32921733

R18 B.Tech. AI&ML DEEP LEARNING LAB JNTU Hyderabad

neural network using TensorFlow/Keras and PyTorch:

 Using TensorFlow/Keras:
import tensorflow as tf
from tensorflow.keras import layers
# Create a simple neural network
model = tf.keras.Sequential([
layers.Dense(64, activation='relu',
input_shape=(784,)),
layers.Dense(10, activation='softmax')
])
# Compile the model
model.compile(optimizer='adam',
loss='sparse_categorical_crossentropy',
metrics=['accuracy'])
# Train the model (example data used here)
# model.fit(train_data, train_labels, epochs=10,
batch_size=32)
 Using PyTorch:
import torch
import torch.nn as nn
import torch.optim as optim
# Create a simple neural network
class SimpleNet(nn.Module):
def __init__(self):
super(SimpleNet, self).__init__()
self.fc1 = nn.Linear(784, 64)
self.fc2 = nn.Linear(64, 10)
def forward(self, x):
x = torch.relu(self.fc1(x))
x = self.fc2(x)
return x
model = SimpleNet()
# Define loss function and optimizer
criterion = nn.CrossEntropyLoss()
optimizer = optim.Adam(model.parameters(), lr=0.001)
# Train the model (example data used here)
# for epoch in range(10):
# running_loss = 0.0
# for data, labels in train_loader:
# optimizer.zero_grad()
# outputs = model(data)
# loss = criterion(outputs, labels)

Downloaded by sravanthi JNTUM ([email protected])

 lOMoARcPSD|32921733

R18 B.Tech. AI&ML DEEP LEARNING LAB JNTU Hyderabad

# loss.backward()
# optimizer.step()
# running_loss += loss.item()
# print(f"Epoch {epoch+1}, Loss:
{running_loss/len(train_loader)}")
In these examples, we've created simple neural network architectures using
TensorFlow/Keras and PyTorch, but you can build more complex models
depending on your specific tasks and requirements.
VIVA QUESTIONS:
 How can you install Keras, TensorFlow, and PyTorch libraries on
your system, and what are the differences between these deep
learning frameworks?

 Explain the process of building a simple neural network using Keras

for image classification, and how do you compile and train the
model?

Downloaded by sravanthi JNTUM ([email protected])

 lOMoARcPSD|32921733

R18 B.Tech. AI&ML DEEP LEARNING LAB JNTU Hyderabad

3. Applying the Convolution Neural Network on computer vision

problems
AIM: Applying the Convolution Neural Network on computer vision problems
DESCRIPTION:
CNNs (Convolutional Neural Networks) revolutionize computer vision by
learning spatial hierarchies automatically.
Here are some common computer vision problems where CNNs are frequently
used:
Image Image classification involves assigning a label or category to
Classificatio an input image from a predefined set of classes. CNNs excel
n at this task due to their ability to learn hierarchical features
and capture patterns in images.
Object Object detection involves identifying and localizing multiple
Detection objects of interest within an image. CNNs can be combined
with techniques like Region Proposal Networks (RPNs) or
Single Shot Multibox Detector (SSD) to accomplish object
detection tasks.
Semantic Semantic segmentation involves classifying each pixel of an
Segmentatio image into a specific category. CNNs with fully
n convolutional architectures, such as U-Net or DeepLab, are
commonly used for semantic segmentation.
Instance Instance segmentation goes beyond semantic segmentation
Segmentatio by not only classifying each pixel but also distinguishing
n individual object instances. CNNs, combined with methods
like Mask R-CNN, are often used for instance segmentation
tasks.
Image CNNs can be used for image generation tasks, such as
Generation generating realistic images from scratch. Generative
Adversarial Networks (GANs) and Variational Autoencoders
(VAEs) are popular architectures for image synthesis.
Style Style transfer involves transferring the visual style of one
Transfer image to another while preserving the content. CNNs are
used to separate and manipulate the content and style
representations of the images.
Image Image captioning combines computer vision and natural
Captioning language processing to generate textual descriptions for

Downloaded by sravanthi JNTUM ([email protected])

 lOMoARcPSD|32921733

R18 B.Tech. AI&ML DEEP LEARNING LAB JNTU Hyderabad

input images. CNNs are used to extract visual features,

which are then combined with recurrent neural networks
(RNNs) to generate captions.
Super- Super-resolution aims to upscale low-resolution images to
Resolution higher resolutions. CNNs, particularly architectures like
SRGAN (Super-Resolution Generative Adversarial
Network), are commonly used for super-resolution tasks.
Face CNNs have demonstrated remarkable performance in face
Recognition recognition tasks, where the goal is to identify individuals
based on their facial features.
Image CNNs can be applied to remove noise, artifacts, or restore
Denoising damaged images, making them useful in image denoising
and and restoration tasks.
Restoration

To apply CNNs to these computer vision problems, you'll need labeled datasets
for training, and you'll build and train CNN architectures using popular deep
learning libraries like TensorFlow, Keras, or PyTorch. The specific architecture
and training process will vary depending on the problem at hand. It's also
common to use pre-trained CNN models (e.g., VGG, ResNet, MobileNet) and
fine-tune them on specific tasks if you have limited data.
Overall, CNNs have revolutionized computer vision and continue to advance
the state-of-the-art in a wide range of applications, bringing significant
improvements to the accuracy and robustness of computer vision systems.

VIVA QUESTIONS:
 What is the main advantage of using Convolutional Neural Networks
(CNNs) over traditional fully connected neural networks for computer
vision tasks?
 Explain the typical architecture of a Convolutional Neural Network used
for image classification, and what are the functions of different layers in
this architecture?

Downloaded by sravanthi JNTUM ([email protected])

 lOMoARcPSD|32921733

R18 B.Tech. AI&ML DEEP LEARNING LAB JNTU Hyderabad

4. Image classification on MNIST dataset (CNN model with Fully

connected layer)
AIM: Image classification on MNIST dataset (CNN model with Fully
connected layer)
DESCRIPTION:
Image classification on the MNIST dataset is a classic example and a great
starting point for understanding how to build a Convolutional Neural Network
(CNN) with a Fully Connected Layer for image recognition. The MNIST
dataset consists of 28x28 grayscale images of handwritten digits from 0 to 9.
Each image is associated with a label representing the digit it represents.
Here's a step-by-step guide to building a CNN model with a Fully Connected
Layer for image classification on the MNIST dataset using TensorFlow and
Keras:
Step 1: Import the necessary libraries
import numpy as np
import tensorflow as tf
from tensorflow.keras import layers, models, datasets
Step 2: Load and preprocess the MNIST dataset
# Load MNIST dataset
(train_images, train_labels), (test_images, test_labels) =
datasets.mnist.load_data()
# Normalize pixel values to range [0, 1]
train_images, test_images = train_images / 255.0, test_images / 255.0
# Expand dimensions to add channel dimen
sion (for grayscale images)
train_images = np.expand_dims(train_images, axis=-1)
test_images = np.expand_dims(test_images, axis=-1)
Step 3: Build the CNN model

Downloaded by sravanthi JNTUM ([email protected])

 lOMoARcPSD|32921733

R18 B.Tech. AI&ML DEEP LEARNING LAB JNTU Hyderabad

model = models.Sequential()
# Convolutional layers
model.add(layers.Conv2D(32, (3, 3), activation='relu', input_shape=(28, 28, 1)))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(64, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(64, (3, 3), activation='relu'))
# Flatten the 3D output to 1D
model.add(layers.Flatten())
# Fully Connected layers
model.add(layers.Dense(64, activation='relu'))
model.add(layers.Dense(10, activation='softmax')) # 10 output classes (0-9
digits)
Step 4: Compile the model
model.compile(optimizer='adam',
loss='sparse_categorical_crossentropy',
metrics=['accuracy'])
Step 5: Train the model
model.fit(train_images, train_labels, epochs=5, batch_size=64,
validation_split=0.1)
Step 6: Evaluate the model on the test set
test_loss, test_accuracy = model.evaluate(test_images, test_labels)
print("Test accuracy:", test_accuracy)
Step 7: Make predictions on new data
predictions = model.predict(test_images[:5])
predicted_labels = np.argmax(predictions, axis=1)
print("Predicted labels:", predicted_labels)
print("True labels:", test_labels[:5])

Downloaded by sravanthi JNTUM ([email protected])

 lOMoARcPSD|32921733

R18 B.Tech. AI&ML DEEP LEARNING LAB JNTU Hyderabad

This code will create a CNN model with three convolutional layers followed by
a fully connected layer. It will then train the model on the MNIST dataset and
evaluate its accuracy on the test set. Finally, it will make predictions on five test
images and display the predicted labels along with the true labels.
You can experiment with the number of layers, units, and other hyperparameters
to improve the model's performance. Additionally, you can use techniques like
data augmentation and dropout to further enhance the model's generalization
capabilities.
VIVA QUESTIONS:
 What is the MNIST dataset, and why is it commonly used for testing
image classification models?
 Describe the architecture of a CNN model with a fully connected
layer for image classification on the MNIST dataset.

10

Downloaded by sravanthi JNTUM ([email protected])

 lOMoARcPSD|32921733

R18 B.Tech. AI&ML DEEP LEARNING LAB JNTU Hyderabad

5. Applying the Deep Learning Models in the field of Natural Language

Processing
AIM: Applying the Deep Learning Models in the field of Natural Language
Processing
DESCRIPTION:
Deep learning models have made significant contributions to the field of
Natural Language Processing (NLP), enabling the development of powerful
language models and applications. Some of the key deep learning models used
in NLP include:
Recurrent Neural Networks (RNNs):
RNNs are designed to handle sequential data, making them well-suited for
natural language processing tasks where the order of words matters. They
process input data step-by-step while maintaining hidden states to capture
context. However, traditional RNNs suffer from vanishing gradient problems.
Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) are
popular variants of RNNs that address this issue.
Applications:
Text Sentiment Named Entity Language
classification analysis Recognition modeling
(NER)

Transformers:
Transformers introduced the attention mechanism, enabling more efficient and
parallelized processing of sequential data. They have revolutionized NLP by
capturing long-range dependencies effectively and have become the backbone
of modern language models.
Applications:
Machine Translation (e.g., Text generation Question-Answering
Google's Transformer- (e.g., OpenAI's GPT (e.g., Google's BERT-
based model "BERT" for series) based model "BERT" for
NMT) QA)

Bidirectional Encoder Representations from Transformers (BERT):

BERT is a transformer-based language model pre-trained on a large corpus of
text data. It learns contextualized word representations, allowing it to
understand the context in which a word appears in a sentence. BERT's pre-

11

Downloaded by sravanthi JNTUM ([email protected])

 lOMoARcPSD|32921733

R18 B.Tech. AI&ML DEEP LEARNING LAB JNTU Hyderabad

trained representations can be fine-tuned for a wide range of NLP tasks, making
it a versatile and powerful model.
Applications:
Text Named Entity Sentiment Question-
classification Recognition analysis Answering
(NER)

Generative Pre-trained Transformer (GPT):

GPT is a family of transformer-based language models developed by OpenAI.
GPT-3, in particular, is one of the largest language models ever created, with
175 billion parameters. It demonstrates impressive capabilities in natural
language understanding and generation.
Applications:
Text completion Text generation (e.g., Language translation
creative writing, code
generation)

Convolutional Neural Networks for NLP:

Although more commonly used for computer vision, CNNs can be adapted to
NLP tasks. They are often employed for text classification and sentiment
analysis by treating text as a 1D sequence of tokens.
Applications:
Text classification Sentiment analysis

Sequence-to-Sequence Models:
Sequence-to-sequence models use encoder-decoder architectures to handle tasks
that involve transforming one sequence into another, such as machine
translation and summarization.
Applications:
Machine translation Text summarization

Attention Mechanisms:
Attention mechanisms are not models themselves, but they have been
instrumental in improving the performance of various NLP models. They allow
models to focus on specific parts of the input during processing, enhancing their
understanding and performance.
Applications:
Language translation Text generation
12

Downloaded by sravanthi JNTUM ([email protected])

 lOMoARcPSD|32921733

R18 B.Tech. AI&ML DEEP LEARNING LAB JNTU Hyderabad

These deep learning models, along with their variants and combinations, have
driven significant advancements in the field of Natural Language Processing,
enabling the development of more sophisticated and context-aware language
models and applications. As research progresses, we can expect further
improvements and innovations in NLP, ultimately leading to more accurate and
human-like language processing systems.

VIVA QUESTIONS:
 What is the role of Word Embeddings in Natural Language
Processing, and how do they benefit deep learning models for NLP
tasks?
 How do Recurrent Neural Networks (RNNs) handle sequential data
in Natural Language Processing, and what are some of the challenges
they face?

13

Downloaded by sravanthi JNTUM ([email protected])

 lOMoARcPSD|32921733

R18 B.Tech. AI&ML DEEP LEARNING LAB JNTU Hyderabad

6. Train a sentiment analysis model on IMDB dataset, use RNN layers

with LSTM/GRU notes
AIM: Train a sentiment analysis model on IMDB dataset, use RNN layers with
LSTM/GRU notes
DESCRIPTION:
To train a sentiment analysis model on the IMDB dataset using RNN layers with
LSTM/GRU units, we'll use TensorFlow and Keras. The IMDB dataset contains
movie reviews labeled as positive or negative sentiments.
Step 1: Import the necessary libraries.
import tensorflow as tf
from tensorflow.keras.datasets import imdb
from tensorflow.keras.preprocessing.sequence import pad_sequences
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Embedding, LSTM, GRU, Dense
Step 2: Load and preprocess the IMDB dataset.
# Load the IMDB dataset, keeping only the top 10,000 most frequent words
num_words = 10000
(train_data, train_labels), (test_data, test_labels) =
imdb.load_data(num_words=num_words)
# Pad the sequences to ensure they all have the same length
max_length = 250
train_data = pad_sequences(train_data, maxlen=max_length)
test_data = pad_sequences(test_data, maxlen=max_length)
Step 3: Build the RNN model with LSTM or GRU layers.
embedding_dim = 100
model = Sequential()
model.add(Embedding(input_dim=num_words, output_dim=embedding_dim,
input_length=max_length))
# Choose LSTM or GRU layer
# LSTM Layer
# model.add(LSTM(units=64, dropout=0.2, recurrent_dropout=0.2))
# GRU Layer
model.add(GRU(units=64, dropout=0.2, recurrent_dropout=0.2))
model.add(Dense(units=1, activation='sigmoid'))
Step 4: Compile and train the model.
model.compile(optimizer='adam', loss='binary_crossentropy',
metrics=['accuracy'])
batch_size = 128
14

Downloaded by sravanthi JNTUM ([email protected])

 lOMoARcPSD|32921733

R18 B.Tech. AI&ML DEEP LEARNING LAB JNTU Hyderabad

epochs = 5
model.fit(train_data, train_labels, batch_size=batch_size, epochs=epochs,
validation_split=0.2)
Step 5: Evaluate the model on the test set.
test_loss, test_accuracy = model.evaluate(test_data, test_labels)
print("Test accuracy:", test_accuracy)
That's it! With these steps, you have built and trained a sentiment analysis
model on the IMDB dataset using RNN layers with LSTM/GRU units. The
model will learn to predict whether movie reviews are positive or negative
based on the given text. You can experiment with different hyperparameters,
such as the number of LSTM/GRU units, the embedding dimension, or the
number of epochs, to see how they affect the model's performance.

VIVA QUESTIONS:

 What is the IMDB dataset, and what is the goal of training a

sentiment analysis model on this dataset using RNN layers with
LSTM/GRU?

 Explain the role of LSTM/GRU units in the sentiment analysis model

and how they address the vanishing gradient problem in traditional
RNNs?

15

Downloaded by sravanthi JNTUM ([email protected])

 lOMoARcPSD|32921733

R18 B.Tech. AI&ML DEEP LEARNING LAB JNTU Hyderabad

7. Applying the Autoencoder algorithms for encoding the real-world data

AIM: Applying the Autoencoder algorithms for encoding the real-world data
DESCRIPTION:
Autoencoders are a class of neural network architectures used for unsupervised
learning. They are particularly useful for encoding real-world data into a lower-
dimensional representation, also known as the "latent space" or "encoding
space." Autoencoders consist of two main parts: the encoder, which maps the
input data to the latent space, and the decoder, which reconstructs the data from
the encoded representation. The objective is to minimize the difference between
the original input and the reconstructed output, encouraging the model to learn a
compact representation of the data.
Applying Autoencoders for encoding real-world data involves the following
steps:

Step 1: Data Preparation

Collect and preprocess the real-world data you want to encode. Ensure that the
data is in a suitable format and normalize it if necessary.

Step 2: Build the Autoencoder Model

Create the Autoencoder model using a neural network library like TensorFlow
or Keras.
import tensorflow as tf
from tensorflow.keras.layers import Input, Dense
from tensorflow.keras.models import Model
# Define the input size
input_size = <your_input_size>
# Define the size of the latent space (encoded representation)
latent_dim = <your_desired_latent_dimension>
# Encoder
input_data = Input(shape=(input_size,))
encoded = Dense(latent_dim, activation='relu')(input_data)
# Decoder
decoded = Dense(input_size, activation='sigmoid')(encoded)
# Autoencoder model
autoencoder = Model(input_data, decoded)

Step 3: Compile and Train the Autoencoder

Compile the model with an appropriate optimizer and loss function, and then
train the Autoencoder on your real-world data.
autoencoder.compile(optimizer='adam', loss='mean_squared_error')
# Assuming you have your data loaded into 'data'
16

Downloaded by sravanthi JNTUM ([email protected])

 lOMoARcPSD|32921733

R18 B.Tech. AI&ML DEEP LEARNING LAB JNTU Hyderabad

autoencoder.fit(data, data, epochs=50, batch_size=32)

Step 4: Encode Real-World Data

Once the Autoencoder is trained, you can use the encoder part of the model to
encode your real-world data into the latent space.
encoder = Model(input_data, encoded)
encoded_data = encoder.predict(data)
The encoded_data variable now contains the encoded representation of your
real-world data in the lower-dimensional latent space.

Step 5: Decode Encoded Data (Optional)

If desired, you can use the decoder part of the model to reconstruct the data
from the encoded representation.
# Assuming you have the encoder part of the model stored in 'encoder'
decoder_input = Input(shape=(latent_dim,))
decoded_output = autoencoder.layers[-1](decoder_input) # Get the last layer of
the decoder
decoder = Model(decoder_input, decoded_output)
# Reconstruct data from encoded representation
reconstructed_data = decoder.predict(encoded_data)

Autoencoders are powerful tools for dimensionality reduction and feature

learning in real-world data. They are widely used in various applications,
including anomaly detection, data compression, and denoising. By training an
Autoencoder, you can obtain a more compact representation of your data, which
can be useful for downstream tasks or visualizations.

VIVA QUESTIONS:

 What is the main purpose of using autoencoder algorithms in

encoding real-world data, and how do they work?

 Explain the typical architecture of an autoencoder model, and what

are some popular variations of autoencoders used for different real-
world data encoding tasks?

17

Downloaded by sravanthi JNTUM ([email protected])

 lOMoARcPSD|32921733

R18 B.Tech. AI&ML DEEP LEARNING LAB JNTU Hyderabad

8. Applying Generative Adversial Networks for image generation and

unsupervised tasks.
AIM: Applying Generative Adversial Networks for image generation and
unsupervised tasks.
DESCRIPTION:
Generative Adversarial Networks (GANs) are a powerful class of deep learning
models used for various tasks, including image generation and unsupervised
learning. GANs consist of two neural networks: a generator and a discriminator.
The generator generates fake data (e.g., images), while the discriminator tries to
distinguish between real data and fake data generated by the generator. The two
networks are trained simultaneously in a competition, where the generator tries
to produce data that fools the discriminator, and the discriminator tries to get
better at distinguishing real from fake data.
Here's how GANs can be applied for image generation and unsupervised tasks:
Image Generation with GANs:
The primary application of GANs is image generation. The generator learns to
create new images that resemble the training data. The discriminator provides
feedback to the generator, guiding it to produce more realistic images over time.

Step 1: Import the necessary libraries.

import numpy as np
import tensorflow as tf
from tensorflow.keras.layers import Dense, LeakyReLU, BatchNormalization,
Reshape, Conv2DTranspose, Conv2D, Flatten
from tensorflow.keras.models import Sequential

Step 2: Build the Generator and Discriminator models.

def build_generator(latent_dim):
model = Sequential()
model.add(Dense(256, input_dim=latent_dim))
model.add(LeakyReLU(alpha=0.01))
model.add(BatchNormalization(momentum=0.8))
model.add(Dense(512))
model.add(LeakyReLU(alpha=0.01))
model.add(BatchNormalization(momentum=0.8))
model.add(Dense(1024))
model.add(LeakyReLU(alpha=0.01))
model.add(BatchNormalization(momentum=0.8))
model.add(Dense(28 * 28 * 1, activation='tanh'))
model.add(Reshape((28, 28, 1)))
18

Downloaded by sravanthi JNTUM ([email protected])

 lOMoARcPSD|32921733

R18 B.Tech. AI&ML DEEP LEARNING LAB JNTU Hyderabad

return model
def build_discriminator(img_shape):
model = Sequential()
model.add(Flatten(input_shape=img_shape))
model.add(Dense(512))
model.add(LeakyReLU(alpha=0.01))
model.add(Dense(256))
model.add(LeakyReLU(alpha=0.01))
model.add(Dense(1, activation='sigmoid'))
return model

Step 3: Define the GAN model by combining the Generator and

Discriminator.
def build_gan(generator, discriminator):
discriminator.trainable = False
model = Sequential()
model.add(generator)
model.add(discriminator)
return model

Step 4: Train the GAN model.

# Define hyperparameters
latent_dim = 100
img_shape = (28, 28, 1)
# Build and compile the discriminator and generator
generator = build_generator(latent_dim)
discriminator = build_discriminator(img_shape)
gan = build_gan(generator, discriminator)
discriminator.compile(loss='binary_crossentropy',
optimizer=tf.keras.optimizers.Adam(0.0002, 0.5))
gan.compile(loss='binary_crossentropy',
optimizer=tf.keras.optimizers.Adam(0.0002, 0.5))
# Load and preprocess real image data (e.g., from MNIST dataset)
# Training loop
epochs = 10000
batch_size = 64
half_batch = batch_size // 2
for epoch in range(epochs):
# Train the discriminator
idx = np.random.randint(0, X_train.shape[0], half_batch)
19

Downloaded by sravanthi JNTUM ([email protected])

 lOMoARcPSD|32921733

R18 B.Tech. AI&ML DEEP LEARNING LAB JNTU Hyderabad

real_images = X_train[idx]
noise = np.random.normal(0, 1, (half_batch, latent_dim))
generated_images = generator.predict(noise)
d_loss_real = discriminator.train_on_batch(real_images, np.ones((half_batch,
1)))
d_loss_fake = discriminator.train_on_batch(generated_images,
np.zeros((half_batch, 1)))
d_loss = 0.5 * np.add(d_loss_real, d_loss_fake)
# Train the generator
noise = np.random.normal(0, 1, (batch_size, latent_dim))
valid_labels = np.ones((batch_size, 1))
g_loss = gan.train_on_batch(noise, valid_labels)

Unsupervised Tasks with GANs:

GANs can be used for various unsupervised learning tasks, such as data
augmentation, feature extraction, and anomaly detection. For example, a pre-
trained GAN generator can be used to generate additional data for training
purposes, which can improve the performance of other models. GANs can also
be used for feature extraction by using the intermediate layers of the generator
as feature representations for downstream tasks.
In summary, GANs are versatile models that have proven to be effective in
image generation and unsupervised learning tasks. They have a wide range of
applications and continue to be an active area of research in the deep learning
community.

VIVA QUESTIONS:

 How do Generative Adversarial Networks (GANs) work for image

generation, and what are the main components of a GAN model?
 How can GANs be applied to unsupervised tasks, and what are some
examples of unsupervised learning tasks where GANs have been
successfully used?

20

Downloaded by sravanthi JNTUM ([email protected])