National University of Computer and Emerging Sciences

Introduction to Course

AI-4009 Generative AI

Dr. Akhtar Jamil

Department of Computer Science

09/09/2024 Presented by Dr. AKHTAR JAMIL 1

 Instructor Information

Instructor Akhtar Jamil E-mail [email protected]

Post Doctorate : Istanbul University-Cerrahpasa, Istanbul, Turkey

PhD : Yildiz Technical University, Istanbul, Turkey
Research Interest : Machine Learning and Computer Vision

Office C-202 F Tel 111-128-128 Ext. 633

09/09/2024 Presented by Dr. AKHTAR JAMIL 2

 Instructor Information
• Associate Professor
• Research Interests
– Applied Machine Learning / Deep Learning
– Computer Vision
– AI in Medial / Business / Remote Sensing

– halanch599 dr.akhtarjamil
LinkedIn

09/09/2024 Presented by Dr. AKHTAR JAMIL 3

 Meeting Hours

 Schedule and Office Hours (Spring-2024)

8:30 10:00 11:30 13:00 14:20 18:45
Days
09:50 11:20 12:50 14:20 15:50 20:05

Office
Monday Gen AI Gen AI
Hour
Office
Tuesday
Hour

Wednesday Gen AI Gen AI

Office Office
Thursday
Hour Hour

Friday

09/09/2024 Presented by Dr. AKHTAR JAMIL 4

 Course Information
• Generative Artificial Intelligence
• Course Credits: 3 + 0
• Theoretical + Practical
Python, TensorFlow, PyTorch, Keras, FastAI,
PyTorch GAN Zoo, TensorFlow GAN (TF-GAN), and
GANs in Keras, OpenAI's GPT, T5 (Text-To-Text
Transfer Transformer), Microsoft's DeepSpeed
Data Sets:
UCI Machine Learning Repository,
Kaggle,
Google
09/09/2024 Dataset Search Presented by Dr. AKHTAR JAMIL 5
 Course Materials
• Will be shared via Google Classroom
Class Code

d3cvhw3
09/09/2024 Presented by Dr. AKHTAR JAMIL 6
 Text Book
• No Text Book

09/09/2024 Presented by Dr. AKHTAR JAMIL 7

 Reference Materials
• Recent Research Papers
• Deep Learning, by Ian Goodfellow and Yoshua Bengio and
Aaron Courville, MIT Press.
• Neural Networks and Deep Learning : A Textbook, Charu C.
Aggarwal
• Deep Learning with Python, François Cholle
• Machine Learning and Deep Learning with Python, scikit-
learn and TensorFlow, Sebastian Raschka and Vahid
Mirjalili
09/09/2024 Presented by Dr. AKHTAR JAMIL 8
 Grading Policy
• Absolute Grading

09/09/2024 Presented by Dr. AKHTAR JAMIL 9

 Grading Policy

Assessment Item Number Weight (%)

Assignments 3–4 10
Quizzes 3–4 10
Project 1 10
Sessional-I 1 15
Sessional-II 1 15
Final Exam 1 40

09/09/2024 Presented by Dr. AKHTAR JAMIL 10

 Explanation of Assessment
• All assignments and Quizzes carry equal weightage
• Enough Time will be given for Assignments for their
implementation.
• Quizzes can be announced or unannounced.
• Project will be done individually.
– Project include writing a research paper for publication.
– You will be using Overleaf for Writing the paper and presenting the
results.

09/09/2024 Presented by Dr. AKHTAR JAMIL 11

 Submission for Assignment / Project
• Every assignment and project deliverable will have a 2-
hour grace period.
– During which no penalty will apply.
– This is intended to allow you time to verify that your submission
has been submitted properly.
• We recommend you download submitted materials and
look it over to make sure all deliverables have been
answered and properly uploaded.
• Deliverables after the grace period will receive a grade of 0.

09/09/2024 Presented by Dr. AKHTAR JAMIL 12

 Course Plagiarism Policy

• Plagiarism in any kind of assessment including project or sessional/ final

exam, assignments quizzes, will result in F grade in the course.

• So what is it?

Plagiarism is presenting someone else's work or ideas as your

own, with or without their consent, by incorporating it into your
work without full acknowledgement. All published and unpublished
material, whether in manuscript, printed or electronic form, is
covered under this definition.

09/09/2024 Presented by Dr. AKHTAR JAMIL 13

 Course Plagiarism Policy
• Project Paper
– Plagiarism for paper will be checked via Turnitin software
– Any paper with score greater than 20% will be considered plagiarized and
ZERO marks will be awarded.
– Content will be also checked for AI Generated plagiarism, it must not
exceed 20% too.

09/09/2024 Presented by Dr. AKHTAR JAMIL 14

 Missed Assessment
• Retake of missed assessment items (other than sessional/
final exam) is NOT allowed.
• Missed assessment item (other than sessional / final exam)
will earn zero marks
• Late submission will NOT be accepted (Even by one
second).
• For missed sessional/ final exam due procedure will be
followed.
• No change is any deadline
09/09/2024 Presented by Dr. AKHTAR JAMIL 15
 Attendance policy
• Students are supposed to have 100% attendance.
• The minimum attendance requirement at all levels and in
all courses is 80%.
• The relaxation of 20% attendance has been given only to
cover any planned events or unforeseen situations.
• I will take attendance at the start of lecture, within 10
minutes.
• Beyond 15 minutes, you will be marked ABSENT.

09/09/2024 Presented by Dr. AKHTAR JAMIL 16

 Student Responsibility
• Mobile Usage
– Use of mobile phones should be avoided.
– You can leave the class in case of an emergency
• Discussion
– Class participation is highly appreciated
– You can ask questions in the class(Highly
Appreciated)
– Discussion among students is not allowed.

09/09/2024 Presented by Dr. AKHTAR JAMIL 17

 Goals of this Course
• This course is designed to gain theoretical and practical
knowledge of Generative Models.
• Course will cover some of the advanced level algorithms.
• This course is about the algorithmic foundations of generative
models.
• Explain principles why and how each generative modeling
algorithm works.
• You will be focusing on implementation techniques

09/09/2024 Presented by Dr. AKHTAR JAMIL 18

 Paper Writing
• A typical paper must consists of following main section
– Abstract
• Should be one paragraph explaining what is being presented in the paper
– Introduction
• A short explanation of the problem and reviews of the existing research.
• You must download and review at least 10-15 papers (Journal Papers are highly encouraged)
– Data Set
• Which data has been used for this study
– Methodology
• What method(s) has been implemented
– Experimental Results
• Explanation of the results obtained. Should include figures, tables etc.
– Conclusion
• Summarize what you did overall
– References
• Bibliography of which papers were reviewed or used in the paper

09/09/2024 Presented by Dr. AKHTAR JAMIL 19

 Reading Papers
• Every week students must read a research paper related to their
topic and write a short report.
– Your weekly work will directly contribute to your assessment items.
• Particularly following items are required for discussion
– Methodology of the paper
– Main Contributions
– Experimental Results
– Strong points and weak points

09/09/2024 Presented by Dr. AKHTAR JAMIL 20

 Prerequisite(s)
• There is no prerequisite for this course.
• However, having good programming skills in Python / C++ etc, is
beneficial
• Advanced Level of Deep Learning, we assume knowledge of DL
models
• Machine Learning Frameworks
– Scikit-Learn (Conventional ML)
– Tensorflow / Keras , PyTorch, DEEPLEARNING4J, Microsoft Cognitive
Toolkit, Caffe

09/09/2024 Presented by Dr. AKHTAR JAMIL 21

 Prerequisites
• Basic knowledge of Optimization and Probabilities:
– Gradients, gradient-descent optimization, backpropagation
– Random variables, independence, conditional independence
– Bayes rule, chain rule
• Basic knowledge of deep neural networks (ANNs, CNNs, RNNs,
Autoencoders).

09/09/2024 Presented by Dr. AKHTAR JAMIL 22

 Projects
• Course projects will be done by individually and can fall into one
or more of the following categories:
– Application of deep generative models on a novel task/dataset
– Algorithmic improvements into the evaluation, learning and/or inference of
deep generative models
– Theoretical analysis of any aspect of existing deep generative models
• I encourage students to do research to find suitable topics
– We can also suggest few possible projects

09/09/2024 Presented by Dr. AKHTAR JAMIL 23

 Presentation
• Each student will be given 8-10 minutes for their presentation
• Students will explain every step they have performed for
preparing their paper and the results obtained.
• Implementation must be shown to the instructor

09/09/2024 Presented by Dr. AKHTAR JAMIL 24

 Programming resources
• https://www.tensorflow.org/
• https://pytorch.org/
• https://www.nltk.org/
• https://scikit-learn.org/stable/
• https://docs.microsoft.com/en-us/cognitive-toolkit/

09/09/2024 Presented by Dr. AKHTAR JAMIL 25

 High Impact Journals/Conferences
• High Impact Conferences
– https://www.scimagojr.com/journalrank.php?area=1700&type=p&category
=1702

• High Impact Journals

– https://www.scimagojr.com/journalrank.php?category=1702&area=1700&t
ype=j

09/09/2024 Presented by Dr. AKHTAR JAMIL 26

 Outline
Week List of Topics
1. -Courses Overview
- Introduction to Deep Learning
- Generative Models
- Bayesian Learning
2 Introduction to ANNs and CNNs
3 Introduction to Probabilistic Graphical Models
4 Restricted Boltzmann Machines (RBMs)
5 Autoregressive Models in Image Generation
6 Sessional -I

09/09/2024 Presented by Dr. AKHTAR JAMIL 27

 Outline
7 Latent Variable Models
8 Flow-based Models (Normalizing Flows)
9 Introduction to Variational Autoencoders (VAEs)
10 Generative Adversarial Networks (GANs)
- Wasserstein GAN
- GAN in NLP, CV
- e.g Deep Fakes, BigGAN, StyleGAN2, Pix2Pix and
CycleGAN

11 Transformer Models in Data Generation

12 Sessional - II
09/09/2024 Presented by Dr. AKHTAR JAMIL 28
 Outline
13 Large Language Models (e.g., GPT-3, BERT, BART,
DistillBert, DALL-E)
14 Introduction to Diffusion Models
Audio and Speech Generation Models: (e.g. WaveNet
15 and SampleRN, Tacotron)

16 Project Presentations/ Evaluations

17 Final

09/09/2024 Presented by Dr. AKHTAR JAMIL 29

 It’s Your Turn
• To Introduce Yourself
– Your Name
– Educational Background

09/09/2024 Presented by Dr. AKHTAR JAMIL 30

 Lecture #1

09/09/2024 Presented by Dr. AKHTAR JAMIL 31

 Goals

This lecture will cover:

– Development Environment Setup/ Mendeley/Overleaf
– Research Stages
– Why Machine Learning?
– Why Generative Models

09/09/2024 Presented by Dr. AKHTAR JAMIL 32

 Development Environment Setup
• Download and Install Anaconda
– https://www.anaconda.com/
– Individual Edition is Free
• Visual Studio Code
• Conda Installation of Tensorflow:
– https://docs.anaconda.com/free/anaconda/applications/tensorflow/
• PIP Installation
– https://www.tensorflow.org/install/pip#linux

09/09/2024 Presented by Dr. AKHTAR JAMIL 33

 Overleaf
• Use LaTeX for writing
• The easy to use, online, collaborative LaTeX editor.
• Overleaf is used by over nine million students and
academics at 6,800 institutions worldwide
• Work from anywhere
• https://www.overleaf.com

09/09/2024 Presented by Dr. AKHTAR JAMIL 34

 Getting Started with Research
• Step 1. Select a Topic
– Your interests and Skills
– Supervisor / Project
– Availability of resources
• Step 2. Search for Relevant Information
– IEEE Explore
– arXiv
– Springer
– Scopus
– Web of Science
– Google Scholar
09/09/2024 Presented by Dr. AKHTAR JAMIL 35
 Getting Started with Research..
• Step 3. Organize Resources
– Zotero
– EndNote
– Mendeley
– JabRef
– License
– MarginNote
• Step 4. Write Paper
– Outline the Paper
– Take Notes
– Avoid Plagiarism

09/09/2024 Presented by Dr. AKHTAR JAMIL 36

 Inside Davos: Future of AI
• Sam Altman
• profound impact of artificial
intelligence (AI) on society during the
World Economic Forum in Davos
• We’re going to have digital doctors,
digital people, and these digital
people are going to merge, and
there’s going to have to be a level of
trust.

09/09/2024 Presented by Dr. AKHTAR JAMIL 37

 Inside Davos: Future of AI
• There will be jobs: Microsoft CEO Satya
Nadella on how AI will change workplace
– There will be a reduction in jobs as we know
but different jobs will be created need different
skills
• Microsoft became major force in the
proliferation of AI through its
investment in OpenAI.
• AI’s commercial impact could translate far
more broadly.

09/09/2024 Presented by Dr. AKHTAR JAMIL 38

 What is Machine Learning?
• A branch of artificial intelligence that enables systems to learn and
improve from experience without being explicitly programmed.
• Key Components:
– Data: Essential for training models.
– Algorithms: Learn from data to make predictions or decisions.
• Types of Machine Learning:
– Supervised Learning: Learning with labeled data.
– Unsupervised Learning: Learning from unlabeled data to find patterns.
– Reinforcement Learning: Learning by trial and error to achieve a goal.

09/09/2024 Presented by Dr. AKHTAR JAMIL 39

 Deep Learning
• Uses complex neural networks with many layers (deep neural
networks).
• These networks are capable of automatically discovering and
learning patterns in data
– Superior performance on tasks like image and speech recognition.
• Deep learning models can automatically extract and learn
features from raw data, reducing the need for manual feature
extraction.

09/09/2024 Presented by Dr. AKHTAR JAMIL 40

 Machine Learning vs Deep Learning

09/09/2024 Presented by Dr. AKHTAR JAMIL 41

 Machine Learning vs Deep Learning

09/09/2024 Presented by Dr. AKHTAR JAMIL 42

 Artificial Intelligence

09/09/2024 Presented by Dr. AKHTAR JAMIL 43

 What are Generative Models?
• Generative models are a class of statistical models used in machine
learning that focus on the generation of new data instances.
• Generative models used in unsupervised machine learning.
• They try understanding and replicating the complex probability
distributions of training data.
• Capture the essence of data - its patterns, variations, and structure -
and can produce new data that follows the same statistical properties.
• They contrast with discriminative models, which are designed to
distinguish between different kinds of data instances.

09/09/2024 Presented by Dr. AKHTAR JAMIL 44

 Discriminative Vs Generative Model
• Generative Models
• Generative learn the joint probability distribution P(X, Y) , where
X is the input data and Y is the output label.
• They understand how the data is generated in terms of a
probabilistic model.
• These models can generate new data points that is similar to the
training data.
• e.g. Gaussian Mixture Models, Hidden Markov Models, Naive
Bayes, Variational Autoencoders (VAEs), and Generative
Adversarial Networks (GANs).
09/09/2024 Presented by Dr. AKHTAR JAMIL 45
 Discriminative Vs Generative Model
• Discriminative Models
• Discriminative models learn the conditional probability P(Y | X) , which is
the probability of the output label Y given the input data X .
• They focus on the boundary between classes.
• These models are used to distinguish between different types of data
points.
• They do not inherently generate new data points but classify input data
into predefined categories.
• Examples include Logistic Regression, Support Vector Machines,
Decision Trees, and most of the Neural Networks used for classification
tasks.
09/09/2024 Presented by Dr. AKHTAR JAMIL 46
 References
• Chapter 20, Deep Learning, MIT Press, Ian Goodfellow, Yoshua
Bengio, Aaron Courville
• Lecture slides of https://www.cs.cornell.edu/~kuleshov/

09/09/2024 Presented by Dr. AKHTAR JAMIL 47

 Thank You 

09/09/2024 Presented by Dr. AKHTAR JAMIL 48