Semester: VII

GENERATIVE ARTIFICIAL INTELLIGENCE

Category: Professional Core Elective
(Common to AI,CS,IS)
(Theory)
Course Code : 21AI74H1 CIE : 100 Marks
Credits: L: T: P : 3:0:0 SEE : 100 Marks
Total Hours : 45L SEE Duration : 3.00 Hours

Unit-I 9Hrs
Introduction to Generative Deep Learning, Generative Modeling What Is Generative Modeling? Historical
perspective on Generative AI, Generative Versus Discriminative Modeling, Introduction to Large Language
Models (LLMs), Applications of Large Language Models, Limitations and Risks of Large Language Models
Unit II 9Hrs
Variational Autoencoders Introduction, Autoencoders, The Autoencoder Architecture the Encoder, The
Decoder, Joining the Encoder to the Decoder, Analysis of the Autoencoder
Building a Variational Autoencoder The Encoder The Loss Function Analysis of the Variational
Autoencoder Using VAEs to Generate Faces, Training the VAE, Analysis of the VAE, Generating New
Faces, Latent Space Arithmetic, Morphing Between Faces
Unit III 9Hrs
Generative Adversarial Networks Introduction to GAN (GAN), The Discriminator, The Generator
Cycle GAN Overview, The Generators (U-Net) The Discriminators Compiling the Cycle GAN Training the
Cycle GAN Analysis of the Cycle GAN Creating a Cycle GAN to Paint Like Monet the Generators (ResNet)
Analysis of the Cycle GAN.
Neural Style Transfer Content Loss Style Loss Total Variance Loss Running the Neural Style Transfer
Analysis of the Neural Style Transfer Model
Unit -IV 9Hrs
Diffusion Models Introduction Denoising Diffusion Models (DDM), The Flowers Dataset, The Forward
Diffusion Process, The Reparameterization Trick, Diffusion Schedules, the Reverse Diffusion Process.
Energy-Based Models Introduction Energy-Based Models, The MNIST Dataset, The Energy Function
Sampling, Using Langevin Dynamics
Unit -V 9Hrs
Bias and Fairness in Generative AI: Understanding Bias in AI Types of biases (algorithmic, data, societal)
Fairness Metrics Statistical parity, equal opportunity, disparate impact Mitigation Strategies Pre-processing,
in-processing, and post-processing techniques
Ethical Design and Deployment of Generative AI Ethical AI Design Principles Human-centered design,
ethical by design Deployment Challenges Real-world implementation, monitoring, and feedback loops
Responsible AI Frameworks Guidelines and best practices for ethical deployment

Course Outcomes: After completing the course, the students will be able to
CO1 Apply the concepts and principles of Generative Artificial Intelligence to engineering requirements.
CO2 Design and demonstrate proficiency in implementing and training various generative AI models using
modern tools.
CO3 Investigate the need for Generative AI techniques to solve real-world problems in diverse domains.
CO4 Explore advanced topics and research directions in Generative AI and critically evaluate their
potential applications.
CO5 Equip students with the knowledge to identify and address ethical issues in Generative AI, focusing
on fairness, accountability, transparency, and human rights.

Reference Books
1 Generative Deep Learning: Teaching Machines to Paint, Write, Compose, and Play" by David
Artificial Intelligence and Machine Learning
 Foster,2nd Edition, 2023. ISBN: 978-1492041948. Publisher: O'Reilly Media.
Deep Learning" by Ian Good fellow, Yoshua Bengio, and Aaron Courville.2nd Edition 2016, ISBN:
2
978-0262035613. Publisher: MIT Press.
Fairness and Machine Learning: Limitations and Opportunities ; Author(s) Solon Barocas, Moritz
3
Hardt, Arvind Narayanan, 2023, ISBN-10/ASIN: 0262048612, Publisher: MIT Press
"Responsible Artificial Intelligence: How to Develop and Use AI in a Responsible Way" by Virginia
4
Dignum , 1st Edition, 2021,ISBN 9783030303716, Publisher: MIT Press

RUBRIC FOR THE CONTINUOUS INTERNAL EVALUATION (THEORY)

# COMPONENTS MARKS
1. QUIZZES: Quizzes will be conducted in online/offline mode. TWO QUIZZES will be
conducted & Each Quiz will be evaluated for 10 Marks. THE SUM OF TWO QUIZZES 20
WILL BE THE FINAL QUIZ MARKS.
2. TESTS: Students will be evaluated in test, descriptive questions with different complexity

Analyzing, Evaluating, and Creating). THREE tests will be conducted. Each test will be 40
evaluated for 50 Marks, adding upto 150 Marks. FINAL TEST MARKS WILL BE
REDUCED TO 40 MARKS.
3. EXPERIENTIAL LEARNING: Students will be evaluated for their creativity and
practical implementation of the problem. Case study-based teaching learning (10), Program
40
specific requirements (10), Video based seminar/presentation/demonstration (20)
ADDING UPTO 40 MARKS.
MAXIMUM MARKS FOR THE CIE THEORY 100

RUBRIC FOR SEMESTER END EXAMINATION (THEORY)

Q. NO. CONTENTS MARKS
PART A
1 Objective type questions covering entire syllabus 20
PART B
(Maximum of TWO Sub-divisions only)
2 Unit 1 : (Compulsory) 16
3&4 Unit 2 : Question 3 or 4 16
5&6 Unit 3 : Question 5 or 6 16
7&8 Unit 4 : Question 7 or 8 16
9 & 10 Unit 5: Question 9 or 10 16
TOTAL 100

Artificial Intelligence and Machine Learning