Deep generative models are at the forefront of modern probabilistic reasoning, transforming the way we model and synthesize high-dimensional data such as images, text, and speech. By leveraging deep neural networks as powerful function approximators, recent breakthroughs in generative modeling---coupled with advances in stochastic optimization---have enabled scalable and realistic data generation across diverse domains, from photorealistic image synthesis to protein design. In this seminar, we will explore the learning algorithms behind deep generative models, covering key approaches such as: Autoregressive models (e.g., PixelCNN, GPT), Flow-based models, Latent variable models (e.g., VAEs), Diffusion models (e.g., DALL·E 2, Stable Diffusion), Energy-based models, and Neural compression. By the end of this seminar, students will gain a rigorous understanding of deep generative models and their transformative applications in computer vision, natural language processing, and beyond. A strong foundation in probability and linear algebra is expected, and familiarity with Python is highly recommended.