Deep Generative Models in Medicine - CGSI 2023
Offered By: Computational Genomics Summer Institute CGSI via YouTube
Course Description
Overview
Explore deep generative models in medicine through this 32-minute conference talk from the Computational Genomics Summer Institute. Delve into the fundamentals of generative models, their applications in healthcare, and advanced concepts like robustness, interpretability, and missing data imputation. Examine the potential of large language models in biomedical text generation and mining. Investigate diffusion models, their training processes, and the Diffusion ELBO. Learn about multivariate diffusions, automatic mean and covariance calculations, and conditioning techniques. Gain insights into recent advancements in molecular discovery using generative models and their challenges in the biomedical field.
Syllabus
Intro
What is a generative model?
Robustness
Interpretability
Filling in Missing Data
Large Language Models
Generation
Diffusions get good likelihoods
What is a diffusion model?
How do you train a diffusion model?
Diffusion ELBO
What types of corruptions/inference SDES?
Derivation for critically damped langevin
Do differences matter?
Other processes?
A multivariate ELBO
Automatic mean + covariance
Is the generic approach any slower?
The stationary distribution
Automatic multivariate diffusion training
Similar results with a fraction of the parameters
Conditioning in a Diffusion
Reference
Taught by
Computational Genomics Summer Institute CGSI
Related Courses
Machine Learning with Graphs - Fall 2019Stanford University via YouTube Topographic VAEs Learn Equivariant Capsules - Machine Learning Research Paper Explained
Yannic Kilcher via YouTube Generative Models With Domain Knowledge for Weakly Supervised Clustering
Stanford University via YouTube Deep Generative Models and Stable Diffusion: Revolution in Visual Synthesis - Lecture by Björn Ommer
Lennart Svensson via YouTube Toward Brain Computer Interface - Deep Generative Models for Brain Reading
MITCBMM via YouTube