Rajit Rajpal

I am currently a PhD student at the University of Edinburgh in the MAC-MIGS programme fully funded by the Probabilistic AI Hub. My research interests include:

Generative modeling, particularly Diffusion models and EBMs.
High-dimensional inference and optimization.
Applied to a wide variety of problems spanning vision, language, and science.

I am fortunate to be supervised by Benedict Leimkuhler and Nikolay Malkin.

Email / Github / LinkedIn

Research

Does translation invariance improve generalization in Tractable Generative Models?
Rajit Rajpal, Antonio Vergari
MSc Dissertation, Poster at BayesAI Workshop, Jun-Aug 2024
description / pdf / code /

We impose translation invariance in tractable generative models (MADE, PCs) to answer the open question as to whether it enhances generalization or not.

Towards Frugal Zero Shot Diffusion Based Image Restoration
Rajit Rajpal, Marcos Obando, Dolly Chen, Bernardin Amogou
Poster at International workshop on diffusions in machine learning, July 2024
description / poster / code /

As part of the week long PhD Hackathon Workshop on Generative Modeling, we improved upon the DPS algorithm by improving robustness and efficiency. Evaluated our method using a CelebA pre-trained diffusion model on inpainting, super-resolution, and outpainting. Our group placed second in the competition and our poster was presented at the above listed workshop!

Teaching

I have been a teaching assistant for the following course:

Linear Algebra and Differential Equations (MATH 54) - Summer 2021, Fall 2021, Spring 2022 - David Nadler, Alexander Paulin

Linear Algebra for Data Science (STAT 89A) - Spring 2021 - Michael Mahoney, Ben Erichson

Experience

	University of California, Berkeley Aug 2019 - May 2022 B.A. Data Science (Domain Emphasis: Applied Mathematics)
	NASA Ames Research Center - Rhombus Power March 2022 - March 2023 Data Scientist
	University of Edinburgh Aug 2023 - Aug 2028 MSc/PhD in Computational Applied Mathematics

Cloned from here!
Last updated: May 23, 2023