filmov
tv
Diffusion and Score-Based Generative Models
Показать описание
Yang Song, Stanford University
Generating data with complex patterns, such as images, audio, and molecular structures, requires fitting very flexible statistical models to the data distribution. Even in the age of deep neural networks, building such models is difficult because they typically require an intractable normalization procedure to represent a probability distribution. To address this challenge, we consider modeling the vector field of gradients of the data distribution (known as the score function), which does not require normalization and therefore can take full advantage of the flexibility of deep neural networks. I will show how to (1) estimate the score function from data with flexible deep neural networks and efficient statistical methods, (2) generate new data using stochastic differential equations and Markov chain Monte Carlo, and even (3) evaluate probability values accurately as in a traditional statistical model. The resulting method, called score-based generative modeling or diffusion modeling, achieves record performance in applications including image synthesis, text-to-speech generation, time series prediction, and point cloud generation, challenging the long-time dominance of generative adversarial networks (GANs) on many of these tasks. Furthermore, score-based generative models are particularly suitable for Bayesian reasoning tasks such as solving ill-posed inverse problems, yielding superior performance on several tasks in medical image reconstruction.
Generating data with complex patterns, such as images, audio, and molecular structures, requires fitting very flexible statistical models to the data distribution. Even in the age of deep neural networks, building such models is difficult because they typically require an intractable normalization procedure to represent a probability distribution. To address this challenge, we consider modeling the vector field of gradients of the data distribution (known as the score function), which does not require normalization and therefore can take full advantage of the flexibility of deep neural networks. I will show how to (1) estimate the score function from data with flexible deep neural networks and efficient statistical methods, (2) generate new data using stochastic differential equations and Markov chain Monte Carlo, and even (3) evaluate probability values accurately as in a traditional statistical model. The resulting method, called score-based generative modeling or diffusion modeling, achieves record performance in applications including image synthesis, text-to-speech generation, time series prediction, and point cloud generation, challenging the long-time dominance of generative adversarial networks (GANs) on many of these tasks. Furthermore, score-based generative models are particularly suitable for Bayesian reasoning tasks such as solving ill-posed inverse problems, yielding superior performance on several tasks in medical image reconstruction.
1:32:01
Diffusion and Score-Based Generative Models
0:15:20
Score Based Generative Modeling through Stochastic Differential Equations Best Paper | ICLR 2021
0:15:28
What are Diffusion Models?
1:00:40
Jeffrey Fessler - An Introduction to Score Based Generative Models
3:46:15
Tutorial on Denoising Diffusion-based Generative Modeling: Foundations and Applications
![[Open DMQA Seminar]](https://i.ytimg.com/vi/d_x92vpIWFM/hqdefault.jpg)
0:45:27
[Open DMQA Seminar] Score-Based Generative Models and Diffusion Models
1:15:18
Score Based Generative Models - Part 1
0:46:36
Carola-Bibiane Schönlieb - Score based diffusion models for conditional generation
1:07:07
Score-based Generative Modeling of Graphs via the System of Stochastic Differential Equations
0:52:46
Miika Aittala: Elucidating the Design Space of Diffusion-Based Generative Models
1:20:59
Generative Modeling by Estimating Gradients of the Data Distribution - Stefano Ermon
0:47:59
Denoising Diffusion-based Generative Modeling
0:41:38
James Thornton: Score-Based Generative Modeling with Critically-Damped Langevin Diffusion
0:40:10
PR-400: Score-based Generative Modeling Through Stochastic Differential Equations
0:33:52
Score-based generative modeling in latent space (220220)
1:10:55
An Optimal Control Perspective on Diffusion-Based Generative Modeling | Julius Berner
0:56:16
Flow Matching for Generative Modeling (Paper Explained)
0:17:39
How I Understand Diffusion Models
1:05:46
Paper Club with Ben - Score-Based Generative Modeling Through Stochastic Differential Equations
0:36:08
Score based Generative Model
1:01:03
GenAI Diffusion Models: A tutorial on score-based generative models w/ medical imaging applications
1:16:33
Cornell CS 6785: Deep Generative Models. Lecture 12: Score-Based Generative Models
1:19:53
Score Based Generative Models - Part 2
1:30:35
Score based Generative Modelling || Score based Diffusion Model || Score and Langevin Sampling
Комментарии