About me

I am a postdoctoral fellow and Principal Investigator (PI) at University of California Berkeley (UCB) in the Department of Statistics and ICSI. I am working with Michael Mahoney. Before that I completed a PhD in optimization at University of Edinburgh (UoE) under the supervision of Prof. Jacek Gondzio.

My work focuses on large scale optimization, higher order methods and communication avoiding methods. Generally, anything related to numerical methods in optimization. Another line of work that I am interested in is local graph clustering applications and methods as well as statistical models and methods for computational social science.

I am currently on the job market.

Publications

Submitted journal papers

A. Devarakonda, K. Fountoulakis, J. Demmel and M. Mahoney. Avoiding communication in primal and dual block coordinate descent methods. arXiv:1612.04003, 2017.
(arXiv)

K. Fountoulakis, F. Roosta-Khorasani, J. Shun, X. Cheng and M. Mahoney. Variational Perspective on Local Graph Clustering. arXiv:1602.01886, 2017.
(arXiv)

K. Fountoulakis and R. Tappenden. A Flexible Coordinate Descent Method. arXiv:1507.03713, 2017.
(arXiv, Software)

Published conference papers

D. Wang, K. Fountoulakis, M. Henzinger, M. Mahoney and S. Rao. Capacity Releasing Diffusion for Speed and Locality. ICML 2017.
(arXiv)

J. Shun, F. Roosta-Khorasani, K. Fountoulakis and M. Mahoney. Parallel Local Graph Clustering. Proceedings of the VLDB Endowment, 9(12):1041-1052, 2016
(arXiv,Conference)

Published journal papers

K. Fountoulakis, D. Gleich and M. Mahoney. An optimization approach to locally-biased graph algorithms. Proceedings of the IEEE, 105(2):256-272, 2017.
(arXiv,Journal)

K. Fountoulakis, A. Kundu, E.-M. Kontopoulou and P. Drineas. A Randomized Rounding Algorithm for Sparse PCA. ACM Transactions on Knowledge Discovery from Data (TKDD), 11(3). No. 38, 2017.
(arXiv,Journal)

K. Fountoulakis and J. Gondzio. A Second-Order Method for Strongly-Convex L1-Regularization Problems. Mathematical Programming, 156(1-2):189-219, 2016.
(Tech. report, arXiv, Journal, Software)

K. Fountoulakis and J. Gondzio. Performance of First- and Second-Order Methods for L1-Regularized Least Squares Problems. Computational Optimization and Applications, 65(3):605-635, 2016.
(Tech. report, arXiv, Journal, Software)

I. Dassios, K. Fountoulakis and J. Gondzio. A Preconditioner for a Primal-dual Newton Conjugate Gradients Method for Compressed Sensing Problems. SIAM J. Sci. Comput., 37(6):A2783-A2812, 2015.
(Tech. report, arXiv, Journal, Software)

K. Fountoulakis, J. Gondzio, and P. Zhlobich. Matrix-free interior point method for compressed sensing problems. Mathematical Programming Computation, 6(1): 1-31, 2014.
(Tech. report, arXiv, Journal, Software)

Technical reports

D. Zhang, K. Fountoulakis, J. Cao, M. Yin, M. Mahoney and A. Pozdnoukhov. Social Discrete Choice Models. arXiv:1703.07520.
(arXiv)

PhD Thesis

Higher-Order Methods for Large-Scale Optimization. School of Mathematics, University of Edinburgh.
(PDF)

Software

Trillion (Download Trillion)

Instance generators for l1-regularized over- and underdetermined least squares. The generators are implemented in MATLAB, they are memoryless and computationally inexpensive. Hence, large-scale instances can be created.

pdNCG: primal-dual Newton Conjugate Gradients (Download pdNCG)

A MATLAB implementation for the solution of unconstrained l1-regularized problems. For example, Machine Learning problems, such as l1-regularized least-squares and logistic regression, Compressed Sensing problems, such as l1-synthesis, l1-analysis and isotropic total-variation. The solver is memoryless, it requires only matrix-vector product operations, hence it is appropriate for large-scale instances.

MFIPMCS: Matrix-free Interior Point Method for Compressed Sensing (Download MFIPMCS)

An interior point method implemented in MATLAB for the solution of real valued compressed sensing problems. The "matrix-free" implies that only matrix-vector products operations are allowed and the process is memoryless. The solver implements efficient preconditioning techniques for the fast solution of linear systems at every iteration.

FCD: Flexible Coordinate Descent (Dowload MATLAB scripts.)

This software only reproduces the experiments in paper "Flexible Coordinate Descent".

Contact

Address: 1947 Center Street, Ste. 600, Berkeley, CA 94704

Email:

Talks

UCB RISELab Summer Retreat 2017 (Communication Avoiding Optimization Algorithms)
(Download pdf)

Workshop on Distributed and Parallel Data Analysis 2016 (Parallel Local Graph Clustering)
(Download pdf)

Workshop on Algorithms for Modern Massive Data Sets 2016 (Local graph clustering algorithms: an optimization perspective)
(Download pdf, Video)

International Symposium on Mathematical Programming (ISMP2015) (Performance of First- and Second-Order Methods for Large-Scale Optimization)
(Download pdf

Oxford Numerical Analysis Group Internal Seminar (On Newton-type methods for support vector machines and signal reconstruction problems)
(Download pdf)

SIAM Conferecne on Computational Science and Engineering (On primal-dual Newton conjugate gradients for compressed sensing)
(Download pdf)

ERGO Seminar (On robust block coordinate descent and primal-dual Newton conjugate gradients)
(Download pdf)

IMA Conference on the Mathematical Challenges of Big Data 2014 (On robust block coordinate descent)
(Download pdf)

Edinburgh SIAM Student Chapter Conference 2014 (On robust block coordinate descent)
(Download pdf)

4th IMA Conference on Numerical Linear Algebra and Optimisation (On primal-dual Newton conjugate gradients for l1-regularized problems)
(Download pdf)

SIAM Conference on Optimization 2014 (On primal-dual Newton conjugate gradients for sparse signal reconstruction)
(Download pdf)

12th EUROPT Workshop on Advances in Continuous Optimization 2014 (On primal-dual Newton conjugate gradients for l1-regularized problems)
(Download pdf)

20th Conference of the International Federation of Operational Research Societies (IFORS) 2014 (On primal-dual Newton conjugate gradients for sparse signal reconstruction)
(Download pdf)

Computational linear algebra and optimization for the Digital Economy workshop 2013 (On second-order methods for various l1-regularized problems)
(Download pdf)

International Conference on Continuous Optimization (ICCOPT) 2013 (On second-order methods with continuation for various l1-regularized problems)
(Download pdf)

11th EUROPT Workshop on Advances in Continuous Optimization (On second-order methods with continuation for l1-regularized problems in Machine Learning)
(Download pdf)

European Conference on Operational Research EURO/INFORMS 2013 (On second-order methods with continuation for l1-regularized problems in Machine Learning)
(Download pdf)

International Conference on Preconditioning Techniques for Scientific and Industrial Applications 2013 (On preconditioners for second-order methods applied to l1-regularized problems)
(Download pdf)

Conference 3rd IMA Conference on Numerical Linear Algebra and Optimization (NLAO) (On interior point methods for sparse signal reconstruction)
(Download pdf)

Operational Research 54 (OR54) (On interior point methods for sparse signal reconstruction)
(Download pdf)

International Symposium on Mathematical Programming (ISMP2012) (On interior point methods for sparse signal reconstruction)
(Download pdf)

3rd Conference on Optimization Methods and Software (OMS) (On interior point methods for sparse signal reconstruction)
(Download pdf)