Teaching Scheme and Syllabus

For

Bachelor of Technology

In

Ar cial Intelligence

Honors in Ar cial Intelligence

Department of Ar cial Intelligence

Sardar Vallabhbhai Na onal Ins tute of Technology

ti
fi
ti
fi
ti
fi
ti
ti
 Honors in Ar cial Intelligence

Sr. Semester Subject Code Schema Credit

No.
1 4 Advanced Algorithms AI 212 3-1-0 4

2 5 Nature Inspired Compu ng and AI311 3-0-2 4

Op miza on
3 6 Genera ve AI AI312 3-0-2 4
4 7 Advanced Topics in Deep learning AI411 3-1-0 4
ti
ti
ti
ti
fi
ti
 Scheme L T P Credit
B. Tech. II (AI) Semester – IV 3 1 0 04
Advanced Algorithms
AI 212

1. Course Outcomes (COs):

At the end of the course, students will be able to
CO1 recognize the need to study algorithms in terms of time and space
CO2 apply di erent data structures and algorithms to solve given problems.
CO3 ability to solve complex network ow and path op miza on problems using advanced
graph techniques
CO4 u lize computa onal geometry concepts to address real-world problems
CO5 iden fy problems that cannot be solved in polynomial me and develop solu ons with
mathema cal bounds rela ve to the op mal solu on

2. Syllabus

INTRODUCTION (06
Hours)
Review of basic concepts: Sor ng, Searching, Priority Queue, Complexity Analysis,
Divide and Conquer Approach, Backtracking basic fundamental techniques,
Recurrence rela ons, Subs tu on methods
DATA STRUCTURES (07
Hours)
Coun ng Sort, Radix Sort, Lower Bound on Sor ng, Fibonacci Heaps, Red Black Trees,
Segment tree and Fenwick tree
GREEDY ALGORITHMS (04
Hours)
Maximum Bipar te Cover Problem, Maximum Matching in an Unweighted Bipar te
Graph, Egyp an Frac on, Convex Hull Problem, Cats and Rats Problem
DYNAMIC PROGRAMMING (08
Hours)
ti
ti
ti
ff
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
fl
ti
ti
ti
ti
ti
ti
ti
ti
 Bin-packing, Steiner trees, Minimum Edit Distance, Longest Palindromic Subsequence,
Longest Palindromic Substring, Subset Sum Problem, Unique BST, Minimum Par on
ADVANCED GRAPH ALGORITHMS (06
Hours)
Network Flows and its Applica ons: Augmen ng Path, Bo leneck Capacity, Residual
Network, Flow Network, Ford-Fulkerson, Edmonds-Karp, Johnson’s algorithm for
sparse graphs, Max- ow Min-cut Theorem, Longest path problem: General Graph,
Directed Acyclic Graph
COMPUTATIONAL GEOMETRY ALGORITHMS AND THEIR APPLICATIONS (06
Hours)
Visibility Graph, Convex Hull Algorithms, Voronoi Diagram, Sweep Line Algorithm,
Closest Pair of Points Problem, Delaunay Triangula on
NP-COMPLETENESS AND APPROXIMATION ALGORITHMs (08
Hours)
Coping with NP-Hardness, Polynomial- me approxima on, Set Cover Problem, Vertex
Cover Problem, The Subset Sum Problem, Travelling Salesman Problem
Tutorials will be based on the coverage of the above topics separately. (15
Hours)

(Total Contact Time: 45 Hours + 15 Hours = 60 Hours)

3. Tutorials:
1. Computa on of me and space analysis of algorithms.
2. Problems on sor ng and tree
3. Problems on greedy algorithms
4. Problems on dynamic programming
5. Solving problems using advanced graph algorithms
6. Problems on computa onal geometry-based algorithms
7. Providing the proof for NP-completeness problems and solving them

4. Books Recommended:
1. Cormen, Leiserson, Rivest, Stein, “Introduc on to Algorithms”, 3/E, MIT Press, 2009.
2. Ding-Zhu Du , Ker-I Ko , Xiaodong Hu, “Design and Analysis of Approxima on Algorithms”, 1/E,
Springer, 2012.
3. Mark Berg, O ried Cheong, Marc Kreveld, Mark Overmars, “Computa onal Geometry
Algorithms and Applica ons”, 3/E, Springer, 2008.
4. Vijay V. Vazirani, “Approxima on Algorithms”, Springer, 1/E, 2003.
ti
tf
ti
ti
fl
ti
ti
ti
ti
ti
ti
ti
ti
ti
tt
ti
ti
ti
ti
Scheme L T P Credit

B. Tech. II (AI) Semester-V 3 0 2 04

Nature Inspired Computing and Optimization
AI311

CO1 Understand the fundamentals of optimization problems and the principles of nature-
inspired computing techniques, including swam Intelligence and random works
CO2 Analyze and implement popular nature-inspired algorithms such as Particle Swarm
Optimization,
Ant Colony Optimization, and Cuckoo Search to solve complex optimization problems.
CO3 Explore advanced algorithms like Artificial Bee Colony Intelligent Water Drop,
Egyptian Vulture, and Invasive Weed Optimization for application in diverse real-world
domains
CO4 Develop and apply nature-inspired computing methods to address real-life optimization
challenges, including industrial, environmental, and engineering applications.

2. Syllabus

Preliminaries (06hours)

Introduction to Optimization Problems and nature inspired computing, (08hours)

swarm Intelligence, random walks, Particle swarm optimization

Ant colony optimization, Cuckoo search algorithm, bat algorithm (08hours)

Artificial bee colony, Algorithm, intelligent water drop algorithm, (08hours)

Egyptian vulture algorithm
 Invasive weed optimization, glowworm swarm optimization, bacteria (08hours)
foraging flower pollination algorithm, firefly algorithm, dealing with
constraints, multi-objective optimization, applications to real-life

Flower pollination algorithm. firefly algorithm, dealing with constraints, (07hours)

multi-objective optimization, applications to real-life problems
Practicals will be based on the coverage of the above topics separately (30hours)

Total Contact time: 45 Hours + 30 Hours =75Hours

4. Practicals:

1 Implement random walks to explore the search space for solving a basic optimization
problem
2 Solve a benchmark optimization problem (i.e. Sphere Function ) using PSO

3 Implement Cuckoo search algorithm(CSA) for optimizing the Rastrigin function. The
Algorithm should employ Levy Flights for exploration and identify the minimum
values of the Rastrigin function within a specified range

4 Implement a Cuckoo search Algorithm to optimize the sphere function visualize nest
positions and their evaluation over iteration.

5 Develop a Brain storm optimization Algorithm to solve a complex function

optimization problem
6 Implement a Bat Algorithm to optimize a Benachmark function Visualize the position
and velocity update of bats over iterations. Analyze and discuss the performance of the
algorithm.
7 Simulate the flower Pollination Algorithm(FPA) to solve a constrained optimization
problem you will use this algorithm to optimize multivariate functions (e.g. Rastrigin,
Ackley) while handling constraints that limit the feasible search space.

8 Apply the Flower Pollination Algorithm and Firefly Algorithm to solve a real-life
optimization problem such as resource allocation, supply chain optimization, or
production planning.

5. Books Recommended:

1. Discrete Problems in Nature-inspired Algorithms, A. Shukla, and R.Tiwari, CRC Press

Galyor and Francis , 2018
2. Nature-inspired optimization Algorithms, Xin-She Yang, Science Direct 2021
3. Nature-inspired optimization Algorithms, Vasuki. A, Routledge, Taylor and Francis, 2020
 Scheme L T P Credit
3 0 2 04
B. Tech. III (AI) Semester – VI
Genera ve AI
AI312

1. Course Outcomes (COs):

At the end of the course, students will be able to
CO1 Iden fy problems that could be solved using Genera ve AI.
CO2 Understand major components and key concepts of GANS, LLMs, and mul modal systems.
CO3 Understand recent advancements in Genera ve models.
CO4 Analyze and apply genera ve models for image, text, and speech genera on tasks.
CO5 Design applica ons of genera ve models in Pytorch and Keras.

2. Syllabus

Introduc on to Genera ve AI (06 Hours)

Introduc on to Genera ve models, Historical overview of genera ve models, Applica ons of
genera ve AI: image and video genera on, text genera on, speech genera on
Genera ve Adversarial Networks (GANs) (08 Hours)
Overview of GANs, Adversarial training, Variants of GANs: DCGANs, WGANs, WGAN-GP and
LSGANs, Di usion Models, Evalua on Metrics
Applica ons of GANs (06 Hours)
Image genera on, style transfer, text genera on, and data augmenta on.

Large Language Models (LLMs) ( 10 Hours)

The transformer architecture, encoder-decoder and encoder-only models, decoder only
models in LLMs, Pre-training and Fine-tuning LLMs, Genera ve Pre-trained Transformers
(GPT), Prompt Engineering, Prompt Tuning, Instruc on Tuning, Reinforcement Learning using
Applica ons of LLMs (06 Hours)
Text genera on, Machine Transla on, Summariza on, Ques on & Answering, Code
genera on, Text-to-SQL genera on
Exploring Mul -Modal System (04 Hours)
ti
ti
ti
ti
ti
ti
ti
ti
ti
ff
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
 Image and Video Cap oning Systems, Visual Ques on & Answering, Mul -modal LLMs,
Speech-to-Speech Machine Transla on
Ethical Considera ons in Genera ve AI (05 Hours)
Bias and fairness in genera ve models, Misinforma on and deep fakes, Intellectual property
rights and copyright issues, Responsible AI prac ces and guidelines
Prac cals will be based on the coverage of the above topics separately. (30 Hours)

(Total Contact Time: 45 Hours + 30 Hours = 75 Hours)

3. Prac cals:
1. Introduc on to TensorFlow/PyTorch
2. Implemen ng a Basic GAN and advanced GAN techniques
3. Image-to-Image Transla on using Pix2Pix, CycleGAN, StarGAN
4. Fine-tuning a Pre-trained LLM (BERT and its variants, GPT, BART, T5, BERT models for Indian
languages)
5. Applica ons of LLMs: Machine Transla on, Text Summariza on, Ques on & Answering
6. Implemen ng an image cap oning model using a pre-trained vision-language model
7. Integra ng pre-trained ASR, Machine Transla on, and TTS models for developing a Speech-to-
Speech MT model
8. Exploring all the evalua on metrics for all the models
9. Implemen ng Bias Detec on and Mi ga on Models

4. Books Recommended:
1. Deep Learning by Ian Goodfellow, Yoshua Bengio, and Aaron Courville
2. Natural Language Processing with Transformers, Lewis Tunstall, Leandro von Werra,
Thomas Wolf
3. Build a Large Language Model (From Scratch), Sebas an Raschka

5. Reference Books:
1. Josh Pa erson and Adam Gibson, “Deep learning: A prac oner's approach”, O'Reilly
Media, First Edi on, 2017.
2. Online Tutorials: Hugging Face, PyTorch, TensorFlow
3. Latest research papers from top er conferences (ACL, NEURIPS, AAAI, NAACL, COLING,
CONLL, SIGIR, CVPR, ICML, ICLR, ICCV, ECCV)
ti
tt
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
 Scheme L T P Credit
B. Tech. IV (AI ) Semester-VII 3 1 0 4
Advanced Topics in Deep learning
AI411

1. Course Outcomes (COs):

At the end of the course, students will be able to
CO1 Founda onal Understanding of Advanced Deep Learning Concepts.
CO2 Applica on of Genera ve Models such as GAN, VAE, Di usion models.
CO3 Mastery of A en on Models and Transformers.
CO4 3D Computer Vision Techniques tasks such as 3D object detec on, reconstruc on, and
facial analysis.
CO5 Explora on of Cu ng-Edge Research Trends.

2. Syllabus

Exploring Auto encoders (06 Hours)

Autoencoder, Architecture Design of AE, Convolu onal AE, Denoising Auroencoder, U-Net ,
YOLO, Applica ons and Integra on AE, U-Net, and YOLO in real-world tasks.
Deep Genera ve Models (08 Hours)
Genera ve Adversarial Networks (GANs), Varia onal Auto encoders (VAE), Self-A en on
Genera ve Adversarial Network, Progressive GAN, StyleGAN

Di usion models (06 Hours)

Genera ve Modelling Through Denoising, Architecture Design of Di usion Models, Training
of Di usion Models, Denoising Di usion Probabilis c Models, Condi onal di usion models.
Text-to-Image Synthesis (04 Hours)
Latent di usion process, CLIP, Stable Di usion, Imagen, DALL-E, GLIDE, Dream Booth.

Deep Dive into A en on (04 Hours)

A en on Mechanism, Advantages of A en on, Architecture Design of Transformers, Vision
Transformer, Swin-Transformer, a en on in NLP, Reformer , T5 (Text-to-Text Transfer
Transformer)
Deep Learning in 3D Computer Vision (08 Hours)
tt
ff
ff
ti
ti
ti
ti
ti
ti
ti
ff
tt
ti
ti
ti
tt
tti
ti
ti
ti
ff
tt
ff
ti
tt
ti
ti
ti
ti
ff
ti
ti
ff
ff
ti
tt
ti
 Introduc on to 3D Computer Vision, Camera Models and Calibra on, Stereo Vision, Structure
from Mo on (SFM), 3D representa ons (Point cloud, Surface mesh, implicit representa on,
Volumetric representa on), Geometry Processing (Mesh opera on, simpli ca on,
decima on, triangula on), Image forma on (Camera Transforma on, Rendering), 3D
Reconstruc on Techniques, 3D Object Detec on and Recogni on, 3D-CNNs, 3DGANs,
Neural Radiance Fields (NeRF)
Innova ons in Deep Learning (07 Hours)
Deep learning for mul modal systems, Speech recogni on, DeepFake detec on, Adversarial
A ack, Machine transla on, Medical Imaging, Image and Video Analy cs.
Tutorials will be based on the coverage of the above topics (15 Hours)
separately.
(Total Contact Time: 45 Hours + 15 Hours = 60 Hours)

Books Recommended:
1. Dive into Deep Learning: Book by Aston Zhang, Zack C. Lipton, Mu Li, Alex J. Smola.
2. Hartley, R., & Zisserman, A. Mul ple View Geometry in Computer Vision.
3. Ian Goodfellow and Yoshua Bengio and Aaron Courville, Deep Learning, MIT Press
4. Lewis Tunstall, Leandro von Werra, & Thomas Wolf, Natural Language Processing with
Transformers
Reference books:
5. Josh Pa erson and Adam Gibson, “Deep learning: A prac oner' approach”, O'Reilly Media,
1st Edi on, 2017.
6. Anil K. Jain , Arun A. Ross , Karthik Nandakumar , Thomas, Introduc on to Biometrics 2n
edi on, 2025
7. Recent papers and ar cles from IEEE PAMI, ACL, EMNLP, NeurIPS, CVPR, ICCV, and ECCV
conferences

4 Tutorial:
1. Case Study on Deep Learning Research Select a recent research paper from a top- er
conference (NeurIPS, CVPR, ICLR, etc.). Deliverable: A concise case study (3-4 pages).
2. Mathema cal Deriva ons for GAN, VAE, and Di usion Models.
3. Genera ve Models: Compare GANs, VAEs, and Di usion Models in terms of architecture and
applica on domains. Deliverable: Write a report (2-3 pages) including key insights from a
recent paper or study.
4. Theore cal Understanding: Explain the key di erences between Self-A en on and Cross-
A en on in transformer architectures. Deliverable: Submit a wri en explana on (2-3 pages).
5. Hugging Face Resources: h ps://huggingface.co/learn
6. Case Studies on a Mul modal System
7. Top- er conference Paper presenta on (NeurIPS, CVPR, ECCV, ICCV, ICLR, IJCB etc.)
8. Capstone Project
tt
tt
ti
ti
ti
ti
ti
ti
ti
ti
ti
tt
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
ti
tt
ti
ti
ti
ti
ti
ff
ff
ff
ti
ti
ti
tt
ti
ti
ti
ti
ti
ti
tt
ti
ti
ti
fi
ti
ti
ti