About


I am a second-year PhD student in the department of Computer Science at the University of Texas at Austin advised by Adam Klivans. My interests lie at the intersection of Theory and Machine Learning. I am specifically interested in understanding what guarantees we can give for learning deep neural networks.

Prior to this, I received my Bachelors degree form Indian Institute of Technology (IIT) Delhi majoring in Computer Science and Engineering. My bachelor thesis was advised by Parag Singla and Chetan Arora.

Publications


  1. Surbhi Goel, Adam Klivans. "Eigenvalue Decay Implies Polynomial-Time Learnability for Neural Networks." Manuscript 2017. [pdf]
  2. Surbhi Goel, Varun Kanade, Adam Klivans, and Justin Thaler. "Reliably Learning the ReLU in Polynomial Time." Short version accepted at NIPS OPTML Workshop 2016 (Oral Presentation). Full version accepted at COLT 2017. [pdf]
  3. Ishant Shanu, Surbhi Goel, Chetan Arora and Parag Singla. "Exploiting Sum of Submodular Structure for Inference in Very High Order MRF-MAP Problems." Manuscript 2015. [pdf]