MODULE 4 : Unsupervised Learning

Faculty:

Prof. Balaji S. Srinivasan, Wadhwani School of Data Science and Artificial Intelligence, IIT Madras

Concepts Covered :

• Overview of Unsupervised Learning
• Clustering Techniques – K-Means, Hierarchical Clustering, DBSCAN
• Dimensionality Reduction – PCA, t-SNE
• Association Rules
• Anomaly Detection
• Overview of Neural Networks for Unsupervised Learning

Learning Outcomes:

• Primary Outcome – Ability to identify appropriate use-cases for and challenges in various unsupervised algorithms in a business context.
• How unsupervised learning differs from supervised learning.
• Key concepts and terminology in unsupervised learning.
• Understand the basics and applications of different clustering algorithms.
• Understand the concept of dimensionality reduction and why it is important in handling high-dimensional data.
• Learn how dimensionality reduction can be applied to simplify datasets for easier analysis and visualization.
• Understand the concept of association rules and their significance in discovering relationships in data.
• Gain insights into how association rules can drive business strategies like product placement and inventory management.
• Understand what anomaly detection is and its importance in various domains like fraud detection, network security, and industrial defect detection.
• Basic understanding of how neural networks can be applied in unsupervised learning.

MODULE 5 : Deep Learning 1

Faculty:

Prof. Balaji S. Srinivasan, Wadhwani School of Data Science and Artificial Intelligence, IIT Madras

Concepts Covered :

• Overview of deep learning
• Basic principles and terminologies in neural networks.
• The Neural Network Architecture
• The backpropagation algorithm
• Convolutional Neural Networks (CNNs) for Vision.
• Recurrent Neural Networks (RNNs) for Sequential Data
• Key Ideas in Deep Learning: Regularization, Embeddings, Transfer Learning

Learning Outcomes:

• Primary Outcome – Understand when to use Deep Learning for your applications, what architectures
would be most appropriate, and what the various components of a typical Deep Learning pipeline are.
• Develop an understanding of the evolution and fundamentals of deep learning
• Grasp the mechanics of deep neural networks, including the backpropagation process and
various network architectures.
• Understand the role of CNNs in image and visual data processing, including their applications and impact.
• Recognize the importance of RNNs and LSTMs in handling sequential data.
• Gain insights into key deep learning concepts such as regularization, embeddings, and transfer
learning, and understand their significance in practical applications.

MODULE 6 : Deep Learning 2

Faculty:

Prof. Ganapathy Krishnamurthi, Wadhwani School of Data Science and Artificial Intelligence, IIT Madras

Concepts Covered :

Advanced CNN

• State of the Art CNN Architectures
• Image Recognition
• Object Detection

Advanced Sequential Models

• Recurrent Neural Networks (RNNs)
• Attention Mechanism
• Transformer Architecture – Large Language Models

Generative Adversarial Networks (GANs)

Diffusion Models

Learning Outcomes:

• Appreciate practical applications and implications of deep learning in different domains.
• Gain knowledge about various state of the art CNN architectures and their applications like
object detection used in computer vision.
• Understand mechanisms behind image generating models like Dall-E and Stable Diffusion
• Understand transformer architectures that drive LLMs like chatGPT

MODULE 7 : Deployability

Faculty:

Prof. Ganapathy Krishnamurthi, Wadhwani School of Data Science and Artificial Intelligence, IIT Madras

Concepts Covered :

Introduction to AI Deployability

• Challenges in Deploying AI Models
• Lifecycle of a Deployed Model

Model Serving and Management

• APIs
• Cloud Deployment
• Model Versioning

Monitoring and Maintenance

• Continuous Monitoring
• Performance Metrics
• Updating Models

Learning Outcomes:

• Understand the complexities and lifecycle of AI model deployment.
• Learn how to serve and manage AI models in production
• Grasp the importance and methods of maintaining AI models post-deployment.

MODULE 8 : Generative and Responsible AI

Faculty:

Prof. B. Ravindran, Head – Wadhwani School of Data Science and Artificial Intelligence, IIT Madras

Concepts Covered :

• Intro to Generative AI – Large Language Models and Image Generation
• Principles of RAI – Transparency; Accountability; Safety, Robustness and Reliability; Privacy and Security; Fairness and non-discrimination; Human-Centred Values; Inclusive and Sustainable development
• Examples of Ai/ML systems going wrong – bias, robustness, explanations, hallucinations, prompt injection, data leakage, deanonymization, deep fakes, copyright infringement, etc. Examples will be drawn from various incidents
• Examples of tools for RAI – measuring bias/fairness, explanations (Lime/SHAP/Grad cam), audit mechanisms
• Regulation landscape – DPDP act (India), GDPR (EU), EU AI act, US presidential declaration

Learning Outcomes:

• Understand the intuition behind Gen AI models
• Understand the principles of responsible AI
• Appreciate the pitfalls of using AI without proper oversight/auditing mechanism
• Exposure to RAI tools
• Exposure to Data Protection Laws and AI regulations.