Autoregressive Diffusion Models - Machine Learning Research Paper Explained
Offered By: Yannic Kilcher via YouTube
Course Description
Overview
Explore a comprehensive video lecture on Autoregressive Diffusion Models (ARDMs), a novel class of generative models that combines autoregressive and diffusion approaches. Delve into the key concepts, including order-agnostic autoregressive models, discrete diffusion, and their applications in text and image generation. Learn about the efficient training objective, parallel generation capabilities, and the model's adaptability to various generation budgets. Discover how ARDMs outperform discrete diffusion models with fewer steps and their unique suitability for lossless compression tasks. Gain insights into the model's architecture, sampling techniques, and extensions for parallel sampling and depth upscaling. Understand the implications of this research for machine learning and its potential impact on generative modeling and data compression.
Syllabus
- Intro & Overview
- Decoding Order in Autoregressive Models
- Autoregressive Diffusion Models
- Dependent and Independent Sampling
- Application to Character-Level Language Models
- How Sampling & Training Works
- Extension 1: Parallel Sampling
- Extension 2: Depth Upscaling
- Conclusion & Comments
Taught by
Yannic Kilcher
Related Courses
Introduction to Artificial IntelligenceStanford University via Udacity Natural Language Processing
Columbia University via Coursera Probabilistic Graphical Models 1: Representation
Stanford University via Coursera Computer Vision: The Fundamentals
University of California, Berkeley via Coursera Learning from Data (Introductory Machine Learning course)
California Institute of Technology via Independent