Nakul Verma



Tel(212) 853-8471

Nakul Verma studies machine learning and high-dimensional statistics. He focuses on understanding and exploiting the intrinsic structure in data to design effective learning algorithms. His work has produced the first provably correct approximate distance-preserving embeddings for manifolds from finite samples, and has provided improved sample complexity results in various learning paradigms, such as metric learning and multiple-instance learning.

Research Interests

Machine Learning, Algorithms, High-dimensional Statistics, Manifold Learning, and Structure Discovery.

Prior to joining Columbia, Verma worked at the Janelia Research Campus of the Howard Hughes Medical Institute as a research specialist developing statistical techniques to analyze neuroscience data, where he collaborated with neuroscientists to quantitatively analyze social behavior in model organisms using various unsupervised and weakly-supervised machine learning techniques. Verma has also worked at Amazon as a research scientist developing risk assessment models for real-time fraud detection.

Verma received his PhD in 2012 and his BS in 2004, both from the University of California, San Diego.

He was awarded Provost Honors from University of California, San Diego from 2001-2004. He was awarded a Janelia Teaching Fellowship in 2015.

Professional Experience

  • Teaching Faculty, Columbia University, 2017–
  • Research Specialist, Janelia Research Campus, HHMI, 2013–2017
  • Research Scientist, Amazon, 2012–2013
  • Research Intern, Yahoo Labs, 2011
  • Research Intern, Qualcomm, 2008

Professional Affiliations

  • Institute of Electrical and Electronics Engineers (IEEE)
  • International Machine Learning Society (IMLS)

Honors & Awards

  • ICML Reviewer Award, 2015
  • Janelia Teaching Fellowship, 2015
  • Best Paper, Wireless Health, 2012

Selected Publications

  • Samory Kpotufe, Nakul Verma, “Time-accuracy tradeoffs in Kernel prediction: controlling prediction quality”, Journal of Machine Learning Research (JMLR), 2017.
  • Nakul Verma, Kristin Branson, “Sample complexity of learning Mahalanobis distance metrics”, Neural Information Processing Systems (NIPS), 2015.
  • Nakul Verma, “Distance preserving embeddings for general n-dimensional manifolds”, Journal of Machine Learning Research (JMLR), 2013.
  • Nakul Verma, Dhruv Mahajan, Sundararajan Sellamanickam, Vinod Nair, “Learning hierarchical similarity metrics”, Computer Vision and Pattern Recognition (CVPR), 2012.
  • Nakul Verma, “A note on random projections for preserving paths on a manifold”, UC San Diego Tech. Report CS2011-0971, 2011.
  • Boris Babenko, Nakul Verma, Poitr Dollar, Serge Belongie, “Multiple instance learning with manifold bags”, International Conference in Machine Learning (ICML), 2011.
  • Nakul Verma, Samory Kpotufe, Sanjoy Dasgupta, “Which spatial trees are adaptive to intrinsic dimension?”, Uncertainty in Artificial Intelligence (UAI), 2009.
  • Yoav Freund, Sanjoy Dasgupta, Mayank Kabra, Nakul Verma, “Learning the structure of manifolds using random projections”, Neural Information Processing Systems (NIPS), 2007.
  • Sanjoy Dasgupta, Daniel Hsu, Nakul Verma, “A concentration theorem for projections”, Uncertainty in Artificial Intelligence (UAI), 2006.