John Moeller
PhD Student


Book Chapter  Journal  Conference  Workshop  Tech Report]

Journal

2016

  • A Unified View of Localized Kernel Learning. (Project Website)
    By John Moeller,   Sarathkrishna Swaminathan,   Suresh Venkatasubramanian,   
    Vol.abs/1603.01374, CoRR (CORR 2016),  2016.
    Abstract
  • 2012

  • Fast Multiple Kernel Learning With Multiplicative Weight Updates (Project Website)
    By John Moeller,   Parasaran Raman,   Avishek Saha,   Suresh Venkatasubramanian,   
    Vol.abs/1206.5580, CoRR (CORR 2012),  2012.
    Abstract
  • 2009

  • Computing Hulls And Centerpoints In Positive Definite Space (Project Website)
    By P. Thomas Fletcher,   John Moeller,   Jeff M. Phillips,   Suresh Venkatasubramanian,   
    Vol.abs/0912.1580, CoRR (CORR 2009),  2009.
    Abstract
  • Conference

    2016

  • Continuous Kernel Learning. (Project Website)
    By John Moeller,   Vivek Srikumar,   Sarathkrishna Swaminathan,   Suresh Venkatasubramanian,   Dustin Webb,   
    In Proceedings of ECML/PKDD (2) (PKDD 2016),  pages 657-673,  2016.
    Abstract
  • A Unified View of Localized Kernel Learning. (Project Website)
    By John Moeller,   Sarathkrishna Swaminathan,   Suresh Venkatasubramanian,   
    In Proceedings of SDM (SDM 2016),  pages 252-260,  2016.
    Abstract
  • 2014

  • A Geometric Algorithm for Scalable Multiple Kernel Learning. (Project Website)
    By John Moeller,   Parasaran Raman,   Suresh Venkatasubramanian,   Avishek Saha,   
    In Proceedings of AISTATS (AISTATS 2014),  pages 633-642,  2014.
    Abstract
  • 2012

  • Approximate bregman near neighbors in sublinear time: beyond the triangle inequality
    By Amirali Abdullah,    John Moeller,    Suresh Venkatasubramanian
    In Proceedings of ACM Symposium on Computational Geometry (SOCG),  pages 31-40,  June,  2012.
    Abstract
  • 2011

  • Horoball Hulls and Extents in Positive Definite Space
    By P. Thomas Fletcher,    John Moeller,    Jeff M. Phillips,    Suresh Venkatasubramanian
    In Proceedings of 12th Algorithms and Data Structure Symposium (WADS 2011),  pages 386-398,   New York City, New York, USA.,  August ,  2011.
    Abstract

    The space of positive definite matrices P(n) is a Riemannian manifold with variable nonpositive curvature. It includes Euclidean space and hyperbolicspace as submanifolds, and poses significant challenges for the design of algorithms for data analysis. In this paper, we develop foundational geometric structures and algorithms for analyzing collections of such matrices. A key technical contribution of this work is the use of horoballs, a natural generalization of halfspaces for non-positively curved Riemannian manifolds. We propose generalizations of the notion of a convex hull and a center point and approximations of these structures using horoballs and based on novel decompositions of P(n). This leads to an algorithm for approximate hulls using a generalization of extents.