Research Highlights

Improving fairness in the Machine Learning pipeline (2018)

-- Mohit Singh, Uthaipon Tantipongpipat, Santosh Vempala

 

ACO Professors Mohit Singh and Santosh Vempala, along with students Samira Samadi,  Uthaipon Tantipongpipat, and Prof. Jamie Morgenstern, developed a new algorithm to reduce bias and introduce fairness into Machine Learning models. Principal Component Analysis (PCA) is a dimensionality reduction tool ubiquitous to many machine learning algorithms - however it has been shown to inherently add bias to the data. The researchers developed Fair-PCA, an algorithm which eliminates bias and runs as fast as regular PCA.

Their work has been featured in an article by the School of Computer Science at Georgia Tech.


Breakthrough proof of the 40-year old Kelmans-Seymour Conjecture (2016)

-- Dawei He, Yan Wang, Xingxing Yu

 

ACO Professor Yu and students Dawei He and Yan Wang confirmed a famous 40-year old conjecture due (independently) to the world-famous Princeton University graph theorist, Paul Seymour and another expert in discrete mathematics, Alexander Kelmans, based at the University of Puerto Rico. The proof is complex, requiring over 135 manuscript pages, and took more than two years to be verified thoroughly by the experts in the area, and subsequently got published in 2019.

Due to the sensation this impressive breakthrough created, it was featured in an article by Georgia Tech News as well as by several prominent publications:


Breakthrough work on the Cap-Set Problem (2016)

-- Ernie Croot

 

ACO Professor Ernie Croot, along with researchers Vsevolod Lev and Peter Pach presented a significantly improved bound to the Cap-Set Problem, which asks how big a subset of {0,1,2,3}^n can exist that contains no three elements summing to 0. Their proof is an application of the polynomial method, and is simpler than the proofs of previous bounds for the problem.

The result was featured in several popular blogs of famous researchers: