Diffusion Models
Diffusion models (DM) have become the state of the art for sample quality in generative modelling. They work by sequentially corrupting training data with slowly increasing noise and then learning to reverse the corruption. A look at their mathematical foundations reveals how much they borrow from statistical mechanics. By understanding their foundations and recent implementations, practitioners will add a powerful new tool to their ML portfolio.

Research feed

Pill

Planning with Diffusion for Flexible Behavior Synthesis

Diffusion models are finding their way into more machine learning domains. This paper introduces a framework for applying them to …

Fabio Peruzzo
Reinforcement Learning
Diffusion Models
Pill

Prompt-to-Prompt Image Editing with Cross Attention Control

Generating images from text is complicated by the fact that small changes in prompt gives very different results. This new study shows how …

Fabio Peruzzo
Diffusion Models
Pill

Denoising Diffusion and Score Based Generative Models

This pill gives a short overview of the development of diffusion models over the last 7 years. It contains several references useful for …

Fabio Peruzzo
Diffusion Models
Pill

Progressive Distillation for Fast Sampling of Diffusion Models

The generation process of diffusion models can be sped up by using a slow teacher model for training a faster student - the latter …

Fabio Peruzzo
Diffusion Models
Pill

Score-Based Generative Modeling with Critically-Damped Langevin Diffusion

In this rather technical paper, a new dynamics is proposed for the sampling of diffusion models. Instead of sampling the de-noised image …

Fabio Peruzzo
Diffusion Models
Pill

Analytic-DPM: an Analytic Estimate of the Optimal Reverse Variance in Diffusion Probabilistic Models

An analytic form for the de-noising process of diffusion neural networks is found in this important work. Monte Carlo sampling with the …

Fabio Peruzzo
Diffusion Models
Pill

GeoDiff: A Geometric Diffusion Model for Molecular Conformation Generation

Diffusion models are used for predicting the stable 3d conformations of a molecule from its 2d graph.

Fabio Peruzzo
Diffusion Models
Pill

Score-Based Generative Modeling through Stochastic Differential Equations

A seminal contribution to the field of diffusion models, here a connection between de-noising, score-matching and stochastic differential …

Fabio Peruzzo
Diffusion Models

Other series in Advances and fundamentals in ML

Simulation and AI

AI techniques are fundamentally transforming the field of simulation by combining physics-based modeling with data-driven machine learning.

Advances and Fundamentals in ML

Optimization in ML

Optimization is a key component of machine learning. In this series we review recent developments allowing to train larger models, faster …

Advances and Fundamentals in ML

Reinforcement Learning

Recent and popular advances in Reinforcement Learning are known to be data-hungry. Attempts to handle this deficit include developing …

Advances and Fundamentals in ML

Geometric deep learning

Specialized deep learning architectures exploit the intrinsic regularities arising from the underlying structure of the physical world. …

Advances and Fundamentals in ML
Check all of our work