Sumanth Varambally

Sumanth Varambally

Ph.D. student in Data Science at UCSD

Sitemap

A list of all the posts and pages found on the site. For you robots out there is an XML version available for digesting as well.

Pages

Posts

Future Blog Post

less than 1 minute read

Published:

This post will show up by default. To disable scheduling of future posts, edit config.yml and set future: false.

Blog Post number 4

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

Blog Post number 3

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

Blog Post number 2

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

Blog Post number 1

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

portfolio

publications

Generalization on unseen domains via inference-time label-preserving target projections

Published in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2021

We propose a novel method for domain generalization using an inference-time optimization procedure. We first learn a label-sensitive, domain-agnostic metric over images. We then learn a generative model over the resulting representations. Finally, during inference, we optimize over the latent space of the generative model to generate a point from the source domain which preserves the target label.

Recommended citation: Pandey, P., Raman, M., Varambally, S., & Ap, P. (2021). Generalization on unseen domains via inference-time label-preserving target projections. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 12924-12933).

Fast One-class Classification using Class Boundary-preserving Random Projections

Published in Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining, 2021

We develop a novel fast one-class classifier called FROCC. We generate a large number of random directions, and project each point from the training set onto these directions. Along each direction, we group “nearby” projections into intervals. Given a test point, we score it as an inlier or outlier based on the proportion of directions along which it lies in an interval.

Recommended citation: Bhattacharya, A., Varambally, S., Bagchi, A., & Bedathur, S. (2021, August). Fast One-class Classification using Class Boundary-preserving Random Projections. In Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining (pp. 66-74).

Which MAPF Model Works Best for Automated Warehousing?

Published in Proceedings of the International Symposium on Combinatorial Search, 2022

In this paper, we benchmark different problem formulations and algorithms for solving the life-long MAPF problem in a warehouse setting. Our goal is to determine which algorithms and variants have the highest impact on the throughput.

Recommended citation: Varambally, Sumanth, Jiaoyang Li, and Sven Koenig. "Which MAPF Model Works Best for Automated Warehousing?." Proceedings of the International Symposium on Combinatorial Search. Vol. 15. No. 1. 2022.

Finding Patient Zero: Learning Contagion Source with Graph Neural Networks

We are given an interaction graph, whose nodes indicate people, and an edge between two nodes indicates an interaction. We are also given a snapshot, indicating the status of each node as susceptible, infected or recovered. Our goal is to determine the source of the contagion (also known as patient zero). We examine the effectiveness of graph neural networks at this task. We demonstrate the robustness of this method to missing node and edge information. We also examine the effect that the robustness entropy of the interaction graph has on the patient-zero detection accuracy. Check out our poster here.

talks

teaching

MTL100 - Calculus

Undergraduate course, Indian Institute of Technology, Delhi, Mathematics Department, 2020