Computational and Cognitive Vision and Robotics

Algorithms and Systems Analysis

Multimedia Technology

Course Description

Generative models (GMs) are widely used in many subfields of AI, Machine Learning and Data Science. Recent advances in parameterizing these models using deep neural networks, combined with progress in stochastic optimization methods, have enabled scalable modeling of complex, high-dimensional data. In this course, we will study the probabilistic foundations and learning algorithms for deep generative models, including generative adversarial networks (GANs), variational autoencoders (VAEs), normalizing flows (NFs), deep autoregressive (DAR) models, diffusion probabilistic models (DPMs), and energy-based models (EBMs). The course will also discuss application areas that have benefitted from deep generative models, including speech processing, computer vision, reinforcement learning, and inverse problem solving. Demonstrations of deep generative models and code execution will be established. This course includes lectures, research papers, assignments, a student-proposed project and a final exam.
 

Goals

• Expanding the machine/deep learning skills
• Widening the application areas of probability theory
• Design, implement and train deep generative models
• Particularly learn, implement, train and run o   deep autoregressive models
  • variational auto-encoders o      normalizing flows
  • generative adversarial nets o      energy-based models
  • diffusion probabilistic models
• Testing real-life applications such as image/audio generation and domain adaptation