Prem Gopalan
Address: 35 Olden Street
Princeton, NJ 08540, USA
E-Mail: pgopalan@cs.princeton.edu

At Princeton, I was at the Statistical Learning at Princeton Group and the S* Network Systems Group (SNS) in the Computer Science Dept. I was advised by Prof. David Blei and Prof. Mike Freedman.

I've graduated from Princeton. Prior to my PhD, I worked at Mazu Networks (now part of Riverbed Technology) for several years. I received my M.S. in Computer Science from Purdue University.

Research

I develop probabilistic graphical models and approximate posterior inference algorithms to learn latent structures in large data. My research includes fast algorithms for finding overlapping communities and popular nodes in networks, building recommendation systems, discovering population structure in genotype variations, and inference under bayesian nonparametric statistical models of user behavior. Many of these algorithms lie in the variational inference or stochastic variational inference framework. I build efficient open-source software in C++.

Early in my PhD, I was part of the Serval team. I helped build the first version of an end-host stack for a service-centric network architecture.

Working papers

  1. Bayesian inference for Poisson community models
    D. Mimno, P. Gopalan, D.M. Blei
    In submission. Contact authors for manuscript.
  2. Scaling probabilistic models of genetic variation to millions of humans
    P. Gopalan, W. Hao, D.M. Blei, J.D. Storey
    [PDF] [Code]
    bioRxiv preprint

Publications

  1. Thesis: Scalable inference of discrete data: user behavior, networks and genetic variation
    [PDF]
  2. Content-based recommendation with Poisson factorization
    P. Gopalan, L. Charlin, D.M. Blei
    NIPS 2014, Montreal, CA, Dec 2014.
    [PDF] [Code]
  3. Bayesian nonparametric Poisson factorization
    P. Gopalan, F.J.R. Ruiz, R. Ranganath, D.M. Blei
    AISTATS 2014, Reykjavik, IS, April 2014.
    [PDF] [Supplement] [Code]
    Oral presentation
    (Contributed talk at the NIPS Workshop on Probabilistic Models for Big Data, Lake Tahoe, 2013.)
  4. Scalable recommendation with Poisson factorization
    P. Gopalan, J. Hofman, D.M. Blei
    arXiv preprint 2013, arxiv.org/abs/1311.1704
    [PDF] [Code]
  5. Modeling overlapping communities with node popularities
    P. Gopalan, C. Wang, D.M. Blei
    NIPS 2013, Lake Tahoe, NV, Dec 2013
    [PDF] [Code]
  6. Efficient online inference for Bayesian nonparametric relational models
    D. Kim, P. Gopalan, D.M. Blei
    NIPS 2013, Lake Tahoe, NV, Dec 2013
    [PDF]
  7. Efficient discovery of overlapping communities in massive networks
    P. Gopalan, D.M. Blei
    Proceedings of the National Academy of Sciences, 110 (36) 14534-14539, 2013
    [PDF] [Supplement] [Code]
    Direct Submission; Acceptance rate: 19%.
  8. Sampling strategies for the scalable inference of communities
    P. Gopalan, D.M. Blei
    In the NIPS Workshop on Social Network and Social Media Analysis: Methods, Models and Applications, 2012
    Best Student Poster Award
  9. Scalable inference of overlapping communities
    P. Gopalan, D. Mimno, S. Gerrish, M.J. Freedman, and D.M. Blei
    NIPS 2012, Lake Tahoe, NV, Dec 2012
    [PDF] [Code]
    Spotlight Paper (4.9% of submissions)
  10. Serval: An end-host stack for service-centric networking
    E. Nordstrom, D. Shue, P. Gopalan, R. Kiefer, M. Arye, S. Ko, J. Rexford, and M.J. Freedman
    NSDI 2012, San Jose, CA, April 2012
    [PDF]
    Community Award, Honorable Mention