Post Graduate Programme in Machine Learning & AI (Executive)

From Black Box to Glass Box: An Introduction to Explainable AI

• With the increasing adoption of AI, there is a growing need to ensure that AI models can be understood and trusted. This session will introduce you to the concept of Explainable AI and Model Interpretability.
• Objective: To gain a comprehensive understanding of the importance of XAI (Explainable Artificial Intelligence) and model interpretability, and learn how to build and interpret an interpretable machine learning model using LIME & SHAP techniques

Is AI racist and biased?

• As AI is increasingly used in various domains, it is important to consider the ethical implications and potential biases that can arise. This session will introduce learners to the concepts of ethics and bias in AI and provide real-world examples.
• Objective: To understand the ethical and bias considerations in AI and how to identify and mitigate potential issues. Learn through case studies in healthcare, finance, autonomous vehicles, recruitment among others.

From Data to Dialogue: Diving into ChatGPT's Architecture and Functionality

• This session will provide an overview of GPT models and their architecture, along with advanced techniques like RLHF and InstructGPT.
• Objective: The objective of this session is to introduce learners to one of the variant of the Transformer - the GPT model, specifically the Davinci variant. They will gain knowledge about the architecture of GPT models and advanced techniques such as Reinforcement Learning from Human Feedback (RLHF), one-shot & few shot learning. They will also learn how to fine-tune GPT models with prompts and evaluate their performance through hands-on exercises.

ChatGPT-Based Ticket Classifier: A Game-Changer for Support Teams

• Using transfer learning with GPT-3 eliminates the need to train NLP models from scratch, and provides access to state-of-the-art language models on various NLP tasks. This session will help you perform 'Ticket Classification Using ChatGPT'
• Objective: To build a product that can automatically classify support tickets into different categories based on their topics using the fine-tuned ChatGPT model. This will help companies and organizations to improve their support ticket handling efficiency and reduce the workload of their support teams.

Build an Image generator using Stable Diffusion

• Using transfer learning with GPT-3 eliminates the need to train NLP models from scratch, and provides access to state-of-the-art language models on various NLP tasks. This session will help you 'Build an image generator using stable diffusion'
• Objective: The objective of this session is to teach learners how to build an image generator using Stable Diffusion and the Hugging Face library. They will learn how to prepare data, build and train a Stable Diffusion model, evaluate its performance, and build a Streamlit dashboard to showcase the model's output.

The World of Graph Networks

• This session aims to simplify the process and provide learners with a comprehensive understanding of Graph Neural Networks (GNNs) and how to build a simple GNN model.
• Objective: The objective of this session is to introduce learners to the components of a GNN, its working mechanism, and how to build a simple GNN model for a given problem statement. By the end of the session, learners will be able to apply their knowledge to solve problems related to graph data.

Advanced Graph Networks

• With the limitations of simple GNNs, there is a need to explore advanced GNN architectures. This session aims to provide an in-depth understanding of advanced graph networks.
• Objective: By the end of this session, learners should be able to comprehend the drawbacks of simple GNNs. Additionally, learners should be able to compare and evaluate the performance of advanced GNNs like Graph Convolutional Networks(GATs) & Graph Attention Networks(GANs) to simple GNNs

State of Art (STOA) architectures in Computer vision

• In this session, we will introduce learners to the concept of landmark and feature detection, and enable them to build a simple landmark detection system.
• Objective: By the end of the session, learners will be able to understand the fundamentals of landmark detection, popular feature detection algorithms such as SIFT and SURF, and build a landmark detection system using feature detection algorithms.

State of Art(STOA) architectures in generative modelling

• Keeping up with the latest advancements in generative modeling can be challenging. This session aims to provide an overview of the state-of-the-art (STOA) architectures in generative modeling and their applications.
• Objective: By the end of the session, learners should have an understanding of the STOA architectures in generative modeling and be able to read research papers on topics such as image-to-text models (Vision Transformers), text-to-image models(CoCA, CLIP), and multimodality(Pathways).