Generative AI: Taming Stable Diffusions

Although the application date for this project is past, the project is still ongoing. You may still wish to contact this professor about other ways of getting involved with this work. Please attend an info session or contact email for more information.

Diffusion models have become the most successful way in recent past to sample
from a given distribution. One of the most notable implementations is DALL·E from Open AI,
which generates images based on a prompt. In mathematical terms, the problem is to sample from
an underlying distribution from which we only have samples. This projects aims to simulate these algorithms and create a control algorithm for robotics based on these ideas.

 

NSF fuding might be available to pay time invested in the project.

Project Website
Diffusion Models proposal
Qualifications

Qualification: Pro-efficiency in Python.

Preferred Qualifications: Experience in Stochastic Differential Equations and Partial Differential Equations.

Project Timeline

1st month: Understand and Implement the Vanilla stable diffusion algorithm.

2nd month: Prototyping a stable diffusion based control algorithm for robotics.

3rd month: Write report for publication.

Duties

Simulate several related algorithms to Stable Diffusions, and understand hidden biases in the generative process.

Prototype a control algorithm for robotics using Stable Diffusions.

Typical Time Commitment
4
Desired Length of Commitment
1
Tags
machine learning
stochastic processes
Sign in to Bookmark
Project Lead
Who to Contact
Matias Delgadino
Contact Email
matias.delgadino@utexas.edu

I'M INTERESTED IN THIS PROJECT. WHAT SHOULD I DO NEXT?

The Office of Undergraduate Research recommends that you attend an info session or advising before contacting faculty members or project contacts about research opportunities. We'll cover the steps to get involved, tips for contacting faculty, funding possibilities, and options for course credit. Once you have attended an Office of Undergraduate Research info session or spoken to an advisor, you can use the "Who to contact" details for this project to get in touch with the project leader and express your interest in getting involved.

Have you tried contacting professors and need more help? Schedule an appointment for additional support.