Gautam Sreekumar

I am a 4th year Ph.D. student supervised by Dr. Vishnu Naresh Boddeti at Michigan State University. Previously, I have worked at Qualcomm India as a software engineer. I obtained my B.Tech. in Electrical Engineering from Indian Institute of Technology Madras where my undergraduate thesis was advised by Dr. A.N. Rajagopalan.

My research is at the intersection of causality and machine learning. Specifically, I am interested in enforcing causal relations that are derived from domain knowledge or experiments in the representations learned by machine learning models.

Email / Google Scholar / X / Github

Research

	Physics Informed Neural Network for Dynamic Stress Prediction Hamed Bolandi, Gautam Sreekumar, Xuyang Li, Nizar Lajnef, Vishnu Naresh Boddeti Applied Intelligence, 2023 arXiv Enforcing PDE constraints in a model to predict dynamic stress distribution from a sequence of sensor readings of the load on a Gusset plate.
	On the Biometric Capacity of Generative Face Models Vishnu Naresh Boddeti, Gautam Sreekumar, Arun Ross, IJCB, 2023 arXiv Estimating an upper bound on the capacity of generative models to create unique identities using face feature extractors.
	Spurious Correlations and Where to Find Them Gautam Sreekumar, Vishnu Naresh Boddeti, 2nd SCIS Workshop at ICML, 2023 arXiv Jointly analyzing commonly studied causes of spurious correlations in ERM models to aid the design of spurious correlation mitigation techniques.
	Heat-assisted Detection and Ranging Fanglin Bao, Xueji Wang, Shree Hari Sureshbabu, Gautam Sreekumar, Liping Yang, Vaneet Aggarwal, Vishnu Naresh Boddeti, Zubin Jacob Nature, 2023 Paper / Article Press Coverage: Science, Tech Xplore, Popular Science, The Register, New Scientist, Maginative, Phys.org Enabling machine learning models to see in the dark by decomposing passive heat signals into temperature, emissivity and texture.
	Neuro-DynaStress: Preedicting Dynamic Stress Distributions in Structural Components Hamed Bolandi, Gautam Sreekumar, Xuyang Li, Nizar Lajnef, Vishnu Naresh Boddeti ArXiv, 2023 arXiv Predicting dynamic stress distribution from load sequence, geometry and boundary conditions using image-to-image models.
	Neural Architecture Transfer Zhichao Lu, Gautam Sreekumar, Erik Goodman, Wolfgang Banzhaf, Kalyanmoy Deb, Vishnu Naresh Boddeti T-PAMI, 2021 arXiv We provide an evolutionary algorithm to search pareto-optimal architectures from a supernet trained on ImageNet.

Template from Jon Barron.