Nikhil Naik

I am a Prize Postdoctoral Fellow at Harvard University. I obtained my PhD from MIT Media Lab in February 2017, where my advisor was Ramesh Raskar. My research interests are in machine learning, computer vision, and their applications in economics. I have developed algorithms that quantify human perception of visual data at scale (e.g., Streetscore), along with methods to obtain measurements on populations, the built environment, and the economy from millions of geospatial images (e.g., Streetchange). Using these tools, I have studied economic activity and human behavior. I have also developed methods that reduce the need for human expertise and labor in building deep learning systems. Examples include AutoML algorithms for automatic design and training of deep neural networks and optimization techniques to train deep networks with small amounts of data. My industrial collaborations include Google, Microsoft, ExxonMobil, and Samsung.

Email: naik@mit.edu  Address: 400 Main Street, E19-271, Cambridge MA 02142.



Selected Papers

  • Computer Vision Uncovers Predictors of Physical Urban Change, Proceedings of the National Academy of Sciences (PNAS) 2017
  • Designing Neural Network Architectures using Reinforcement Learning, International Conference on Learning Representations (ICLR) 2017
  • Deep Learning the City: Quantifying Urban Perception At A Global Scale, European Conference on Computer Vision (ECCV) 2016
  • Streetscore – Predicting the Perceived Safety of One Million Streetscapes, IEEE Computer Vision & Pattern Recognition (CVPR) Workshops 2014
  • Students Supervised/Co-supervised

    I have been fortunate to advise the following research students at MIT and Harvard:

  • Abhimanyu Dubey, Harvard Post-Baccalaureate Research Fellow (2015–2017). Now a PhD student at MIT.
  • Bowen Baker, MIT Master of Engineering (2016–2017). Now a researcher at OpenAI.
  • Otkrist Gupta, MIT PhD (2016–Present).
  • Karan Dwivedi, IIT Delhi MTech, (2017–Present).
  • These students have been funded through generous grants from Harvard's Star Family Challenge for Promising Scientific Research, Google, and the International Growth Centre.

    Updates

  • (01/18) Paper on accelerating AutoML accepted at ICLR 2018 Workshops.
  • (12/17) Presenting at NIPS 2017 at the workshops on Meta-Learning and Learning with Limited Labeled Data.
  • (10/17) Invited talk on Streetchange at the Imperial College London's Data Science Institute
  • (08/17) Streetchange featured on the Harvard Homepage!
  • (07/17) Our paper on Streetchange appears in the Proceedings of the National Academy of Sciences.
  • (06/17) Invited talk on Streetchange at the Stanford Institute for Economic Policy Research.
  • (05/17) Presented our work on neural network meta-modeling at the New England Machine Learning Day organized by Microsoft Research.