Sanjeev Arora

Sanjeev Arora

Sanjeev Arora will work on providing theoretical foundations for deep learning, including a better understanding of efficiency and provable guarantees. His recent work includes design of algorithms with provable behavior profiles for settings such as deep learning, generative models, natural language processing, and generative adversarial nets. Sanjeev Arora’s research area is Theoretical Computer Science. Specific topics that he has worked on include Computational Complexity (he has written a book on this topic), Probabilistically Checkable Proofs (PCPs), computing approximate solutions to NP-hard problems, geometric embedding of metric spaces, unique games conjecture, complexity of financial derivatives, and provable bounds for Machine Learning. His lab at Princeton University does research on Unsupervised Learning with Provable Guarantees.

  1. S. Arora, R. Ge, Y. Liang, T. Ma, Y. Zhang. Generalization and Equilibrium in Generative Adversarial Nets (GANs). Proc. ICML 2017.
  2. S. Arora, R. Ge, Y. Li, Y. Liang, T. Ma, A. Risteski. A Latent Variable Model Approach to PMI-based Word Embeddings. TACL 2016: 385-399.
  3. Sanjeev Arora, Rong Ge, Ankur Moitra: New Algorithms for Learning Incoherent and Overcomplete Dictionaries. COLT 2014: 779-806
  4. S. Arora, R. Ge, Y. Halpern, D. Mimno, A. Moitra, D. Sontag, Y. Wu, M. Zhu: A Practical Algorithm for Topic Modeling with Provable Guarantees. ICML (2) 2013: 280-288.
  5. S. Arora, A. Bhaskara, R. Ge, T. Ma: Provable Bounds for Learning Some Deep Representations. ICML 2014: 584-592
  6. S. Arora, R. Ge, R. Kannan, A. Moitra. Computing a Nonnegative Matrix Factorization — Provably. Proc. ACM STOC 2012. 145-162.
  7. S. Arora, R. Ge, A. Moitra. Learning Topic Models - Going beyond SVD. Proc. IEEE FOCS 2012. 1-10.
  8. S. Arora, R. Ge, A. Moitra, S. Sachdeva. Provable ICA with Unknown Gaussian Noise, and Implications for Gaussian Mixtures and Autoencoders. NIPS 2012.
  9. S. Arora, R. Ge, S. Sachdeva, G. Schoenebeck: Finding overlapping communities in social networks: toward a rigorous approach. EC 2012: 37-54
  10. S. Arora, Y. Liang, T. Ma. A simple but tough-to-beat baseline for sentence embeddings. ICLR’17 workshop track.