Generative Models - Basics - Lecture 14

Explore the fundamentals of generative models in this 81-minute lecture from MIT's Deep Learning course. Delve into the theoretical foundations and practical applications of various generative modeling approaches, including density models and energy-based models that learn probability distributions over data. Master key sampling methods used to generate new data points from learned distributions. Understand the architecture and training dynamics of Generative Adversarial Networks (GANs), which pit generator and discriminator networks against each other in a competitive learning framework. Learn about autoregressive models that generate data sequentially by modeling conditional probabilities. Discover diffusion models, a powerful class of generative models that learn to reverse a noise corruption process to generate high-quality samples. Gain insights into the mathematical principles underlying each approach and their respective strengths and limitations in different applications.