Generative AI Models Through the Lens of Dense Associative Memory
Institute for Pure & Applied Mathematics (IPAM) via YouTube
Overview
Save Big on Coursera Plus. 7,000+ courses at $160 off. Limited Time Only!
Explore the fascinating world of Dense Associative Memories and their applications in generative AI models through this illuminating lecture by Dmitry Krotov from IBM. Delve into the mathematical framework and intuitive understanding of these modern Hopfield Networks, which boast a significantly larger memory storage capacity than their predecessors. Discover how these powerful tools are revolutionizing AI and neuroscience research. Examine the intriguing connections between Dense Associative Memories and prominent generative AI models, including transformers and diffusion models. Learn about the innovative Energy Transformer, a neural network that seamlessly integrates energy-based modeling, associative memories, and transformers. Gain insights into the emerging perspective that views diffusion models as Dense Associative Memories operating beyond critical memory storage capacity, and understand its implications for analyzing the memorization-generalization transition in these models. Uncover exciting possibilities for future research in this cutting-edge field of AI.
Syllabus
Dmitry Krotov - Generative AI models through the lens of Dense Associative Memory - IPAM at UCLA
Taught by
Institute for Pure & Applied Mathematics (IPAM)