Generative AI : Probabilistic Perspective

About this course

This course, Mastering Generative AI: A Probability Theory Perspective, is a deep theoretical exploration of Generative AI, designed to provide learners with a solid mathematical foundation. The course begins with the fundamentals of probability theory, ensuring that participants develop an intuitive understanding of key probabilistic concepts essential for generative modeling.Building on this foundation, the course systematically introduces various Generative AI architectures, explaining how they function from first principles. Participants will explore Variational Autoencoders (VAEs), Score-Based Models, Energy-Based Models, and Normalizing Flows, all through a probabilistic lens. The course also covers Autoregressive Models such as Transformers, which power modern Large Language Models (LLMs), and Generative Adversarial Networks (GANs), which revolutionized AI-generated media.Unlike many hands-on coding courses, this program emphasizes theoretical rigor, enabling learners to grasp the mathematical underpinnings of these architectures rather than just their implementation. By the end of the course, students will have a deep conceptual understanding of how different generative models work behind the scenes, equipping them with the knowledge to critically analyze and develop their own AI research. Additionally, the course connects these models to their real-world applications, helping learners appreciate their significance in modern AI advancements and future innovations.Overall this course is a novel introduction to Gen AI from the probability point of view.

Course objectives

• to develop a strong mathematical foundation in probability theory
• to explore different generative AI architectures from a theoretical perspective
• to connect generative models to their applications in the real world