About
News
Students
Papers

Grants
Talks



 
Fred (Farbod) Roosta
ARC DECRA Fellow
School of Mathematics and Physics
University of Queensland, Brisbane, Australia
and
International Computer Science Institute, Berkeley, USA

Email: fred.roostauq.edu.au

About

Research Interests

  • Machine Learning
  • Numerical Optimization
  • Randomized Algorithms
  • Computational Statistics
  • Numerical Analysis and Scientific Computing
  • Numerical Linear Algebra
  • Image Processing and Inverse Problems

News

 

  • Feb, 2018:  Our paper on "Out-of-sample extension of graph adjacency spectral embedding" is available on arXiv now.
  • Feb, 2018:  We just posted a paper on arXiv, in which we extend our prior results on non-convex Newton-type methods with inexact Hessian (i.e., this paper), to allow for gradient approximation as well.
  • Feb, 2018:  The code for our paper on GPU-accelerated sub-sampled Newton methods is available here.
  • Feb, 2018:  The code for distributed convex optimization using GIANT is available here.
  • Feb, 2018:  The code for sub-sampled Trust-Region/ARC for large scale non-convex problems is available here.
  • Dec, 2017:  Our paper on accelerated-adaptive gradient based methods, called Flag n' Flare, has been accepted to AISTATS 2018.
  • Dec, 2017:  A short review on different optimization methods for inverse problems is available on arXiv.
  • Nov, 2017:  We just posted a paper on arXiv in which we study some theoretical properties of Rectified MLPs, viewed as kernels, including the invariance of weight distributions as well as some interesting fixed point results.
  • Nov, 2017:  Our paper on establishing a relationship between optimization and local graph clustering just got accepted to Mathematical Programming!

Students

  • Robert Salomone (PhD)
  • Russell (Susumu) Tsuchida (PhD)
  • Rixon Crane (PhD)
  • Yang Liu (MPhil)

 

I always look for enthusiastic MSc/MPhil/PhD students...
If
(a) your research interests lie anywhere in the union of Numerical Optimization, Machine Learning, Randomized Algorithms, Computational Statistics, Scientific Computing, or Numerical Linear Algebra,
&
(b) you are looking to do a MSc/PhD degree in one of the world's leading and highly ranked research universities in beautiful Australia,
then feel free to drop me a line with your current CV and a few words regarding your research/background.

Preprints and Publications

2018

2017

2016

  • Parallel Local Graph Clustering
    Julian Shun, Farbod Roosta-Khorasani, Kimon Fountoulakis and Michael W. Mahoney
    Proceedings of the VLDB Endowment, 9(12), pp. 1041-1052, 2016.

2015

2014

Research Grants

 
  • Efficient Second-Order Optimisation Algorithms for Learning From Big Data
    Program: Discovery Early Career Researcher Award (DECRA)
    Agency: Australian Research Council (ARC)
    Award Period: Starting in 2018 for 3 years
    Role: PI
  • Robust, Efficient, and Local Machine Learning Primitives
    Program: Data Driven Discovery of Models (D3M)
    Agency: DARPA
    Award Period: Starting in 2017 for 4 years
    Role: coPI

Selected Talks

 

  • SIAM Annual Meeting

    Pittsburgh, Pennsylvania, July, 2017

  • Matrix Program on Computational Inverse Problems

    Creswick, Australia, June, 2017

  • SIAM Conference on Optimization

    Vancouver, Canada, May, 2017

  • SIAM Conference on Computational Science and Engineering

    Atlanta, Georgia, February, 2017

  • Computational and Methodological Statistics (CMStatistics)

    Seville, Spain, December, 2016

  • Pacific Institute for the Mathematical Sciences (SCAIM Seminar)

    Vancouver, British Columbia, September, 2016

  • Institute of Applied Mathematics (IAM)

    Vancouver, British Columbia, September, 2016

  • Virginia Tech Math Colloquium

    Blacksburg, Virginia, September, 2016

  • SAMSI Program on Optimization (Invited Plenary Talk)

    Research Triangle Park, North Carolina, August, 2016

  • Recent Advances in Randomized Numerical Linear Algebra

    National Institute of Informatics, Shonan Village Center, Japan, July, 2016

  • Workshop on Algorithms for Modern Massive Data Sets (MMDS)

    Berkeley, California, June, 2016

  • SIAM Conference on Uncertainty Quantification

    EPFL, Lausanne, Switzerland, April, 2016

  • Matrix Computations and Scientific Computing Seminar

    Berkeley, California, March, 2016

  • AMPLab (All Hands Meeting)

    Berkeley, California, September, 2015

  • Sparse Representations, Numerical Linear Algebra, and Optimization

    BIRS, Alberta, October, 2014

  • SIAM Conference on Optimization

    San Diego, California, May, 2014

  • Pacific Institute for the Mathematical Sciences (SCAIM Seminar)

    Vancouver, British Columbia, November, 2013