Dimitris Papailiopoulos

Assistant Professor @ UW-Madison
Electrical and Computer Engineering
Computer Sciences (by courtesy)

I am an assistant professor of Electrical and Computer Engineering and Computer Sciences (by courtesy) at the University of Wisconsin-Madison, a faculty fellow at the Grainger Institute, and a faculty affiliate with the Optimization group at the Wisconsin Institute for Discovery. My research lies in the intersection of machine learning, coding theory, and distributed systems. I am particularly interested in coordination-avoiding algorithms for large-scale machine learning, and in using erasure codes to speed up distributed computation.

Before coming to Madison, I spent two wonderful years as a postdoc at UC Berkeley. At Berkeley, I was a member of the AMPLab and BLISS, and had the pleasure to collaborate with Kannan Ramchandran and Ben Recht on several fun projects. I received my Ph.D. in 2014 from UT Austin, where I was fortunate to be advised by Alex Dimakis. Before UT, I spent 3.5 years as a grad student at USC. Before all that, I received my M.Sc. (2009) and ECE Diploma (2007) from the Technical University of Crete (TUC), located in the beautiful city of Chania.


Curriculum Vitae
Google Scholar

News


Feb. 2017
Invited talk at ITA 2017
Nov. 2016
Invited talk at Asilomar 2016
Oct. 2016
CYCLADES covered by the morning paper
Sep. 2016
SILO Seminar: Speeding-up Large-scale ML Using Graphs and Codes [video]
Aug. 2016
NIPS 2016: CYCLADES: Conflict-free Asynchronous Machine Learning
Authors: X. Pan, M. Lam, S. Tu, D. Papailiopoulos, C. Zhang, M. I. Jordan, K. Ramchandran, C. Re, B. Recht

Abstract: We present CYCLADES, a general framework for parallelizing stochastic optimization algorithms in a shared memory setting. CYCLADES is asynchronous during shared model updates, and requires no memory locking mechanisms, similar to HOGWILD!-type algorithm. Unlike HOGWILD!, CYCLADES introduces no conflicts during the parallel execution, and offers a black-box analysis for provable speedups across a large family of algorithms. Due to its inherent conflict-free nature and cache locality, our multi-core implementation of CYCLADES consistently outperforms HOGWILD!-type algorithms on sufficiently sparse datasets, leading to up to 40% speedup gains compared to the HOGWILD! implementation of SGD, and up to 5x gains over asynchronous implementations of variance reduction algorithms.


Publications

2016

  • CYCLADES: Conflict-free Asynchronous Machine Learning
    X. Pan, M. Lam, S. Tu, D. Papailiopoulos, C. Zhang, M. I. Jordan, K. Ramchandran, C. Re, B. Recht
    NIPS 2016. [arxiv]
  • Speeding-up Distributed Machine Learning Using Codes
    K. Lee, M. Lam, R. Pedarsani, D. Papailiopoulos, K. Ramchandran
    ISIT 2016. [arxiv]
  • On the Worst-Case Approximability of Sparse PCA
    S. O. Chan, D. Papailiopoulos, A. Rubinstein
    COLT 2016. [arxiv]
  • Bipartite Correlation Clustering - Maximizing Agreements
    M. Asteris, A. Kyrillidis, D. Papailiopoulos, A. G. Dimakis
    AISTATS 2016. [arxiv]
  • Locality and Availability in Distributed Storage
    A.S. Rawat, D. Papailiopoulos, A.G. Dimakis, and S. Vishwanath,
    IEEE Transactions on Information Theory. [preprint]

2015

  • Perturbed Iterate Analysis for Asynchronous Stochastic Optimization
    H. Mania, X. Pan, D. Papailiopoulos, B. Recht, K. Ramchandran, M. I. Jordan
    NIPS 2015, OPT Workshop. [long version]
  • Speeding-up Distributed Machine Learning Using Codes
    K. Lee, M. Lam, R. Pedarsani, Dimitris Papailiopoulos, Kannan Ramchandran
    NIPS 2015, ML-Systems Workshop. [arxiv]
  • Parallel Correlation Clustering on Big Graphs
    X. Pan, D. Papailiopoulos, S. Oymak, B. Recht, K. Ramchandran, M. I. Jordan
    NIPS 2015. [long version]
  • Sparse PCA via Bipartite Matchings
    M. Asteris, D. Papailiopoulos, A. Kyrillidis, A. G. Dimakis
    NIPS 2015. [long version]
  • Orthogonal NMF through Subspace Exploration
    M. Asteris, D. Papailiopoulos, A. G. Dimakis
    NIPS 2015. [long version]
  • Optimal Locally Repairable Codes and Connections to Matroid Theory
    I. Tamo, D. Papailiopoulos, and A. G. Dimakis
    IEEE Trans. on Information Theory, 2015. [preprint]

2014

  • Provable Deterministic Leverage Score Sampling
    D. Papailiopoulos, A. Kyrillidis, C. Boutsidis
    KDD 2014. [long version]
  • Finding Dense Subgraphs via Low-Rank Bilinear Optimization
    D. Papailiopoulos, I. Mitlagkas, A. G. Dimakis, and C. Caramanis
    ICML 2014. [short version], [slides], recorded ICML'14 talk given by Yannis.
  • Nonnegative Sparse PCA with Provable Guarantees
    M. Asteris, D. Papailiopoulos, and A. G. Dimakis,
    ICML 2014. [long version], [slides], recorded ICML'14 talk given by Megas.
  • Locality and Availability in Distributed Storage
    A. S. Rawat, D. Papailiopoulos, A. G. Dimakis, and S. Vishwanath
    ISIT 2014. [long version]
  • Locally Repairable Codes
    D. Papailiopoulos and A. G. Dimakis
    IEEE Transactions on Information Theory, September 2014. [IEEEXplore], [arXiv]
  • A Repair Framework for Scalar MDS Codes
    K. Shanmugam, D. Papailiopoulos, A. G. Dimakis, and G. Caire
    IEEE Journal on Selected Areas in Communications, May 2014. [IEEEXplore], [arXiv]
  • The Sparse Principal Component of a Constant-rank Matrix
    M. Asteris, D. Papailiopoulos, and G. N. Karystinos
    IEEE Transactions on Information Theory, March 2014. [IEEEXplore], [arXiv]. MATLAB implementation by Megas.

2013

  • Sparse PCA through Low-rank Approximations
    D. Papailiopoulos, A. G. Dimakis, and S. Korokythakis
    ICML 2013. [long version], [MATLAB code], [slides], recorded ICML'13 talk.
  • Optimal Locally Repairable Codes and Connections to Matroid Theory
    I. Tamo, D. Papailiopoulos, and A. G. Dimakis
    ISIT 2013. [IEEEXplore], [arXiv]
  • XORing Elephants: Novel Erasure Codes for Big Data
    M. Sathiamoorthy, M. Asteris, D. Papailiopoulos, A.G. Dimakis, R. Vadali, S. Chen, and D. Borthakur
    VLDB 2013. project homepage. Press coverage: high scalability, StorageMojo, TechCrunch
  • Repair Optimal Erasure Codes through Hadamard Designs
    D. Papailiopoulos, A. G. Dimakis, and V. R. Cadambe,
    IEEE Transactions on Information Theory, May 2013. [IEEEXplore], [arXiv]
  • Maximum-Likelihood Noncoherent PAM Detection
    D. Papailiopoulos, G. A.-Elkheir, G. N. Karystinos,
    IEEE Transactions on Communications, March 2013. [IEEEXplore], [draft]

2012

  • A Repair Framework for Scalar MDS Codes
    K. Shanmugam, D. Papailiopoulos, A. G. Dimakis, and G. Caire
    Allerton 2012. [IEEEXplore], [arXiv]
  • Locally Repairable Codes
    D. Papailiopoulos and A. G. Dimakis
    ISIT 2012. [IEEEXplore], [arXiv], [slides]
  • Feedback in the K-user Interference channel
    D. Papailiopoulos, Changho Suh, Alexandros G. Dimakis
    ISIT 2012. [IEEEXplore], [long], [slides]
  • Simple Regenerating Codes: Network Coding for Cloud Storage
    D. Papailiopoulos, Jianqiang Luo, Alexandros G. Dimakis, Cheng Huang, and Jin Li
    INFOCOM 2012 (Miniconference). [IEEEXplore], [arXiv], [slides]
  • Maximum-likelihood Blind PAM Detection
    D. Papailiopoulos, G. A.-Elkheir, G. N. Karystinos
    ICC 2012. [IEEEXplore], [long]
  • Interference Alignment as a Rank Constrained Rank Minimization
    D. Papailiopoulos and A. G. Dimakis,
    IEEE Transactions on Signal Processing, August 2012. [IEEEXplore], [arXiv] MATLAB Implementation

2011

  • Repair Optimal Erasure Codes through Hadamard Designs
    D. Papailiopoulos, A. G. Dimakis, and V. R. Cadambe
    Allerton 2011. [IEEEXplore], [long], [slides]
  • Distributed Storage Codes through Hadamard Designs
    D. Papailiopoulos and A. G. Dimakis
    ISIT 2011. [IEEEXplore], [long] [slides]
  • Sparse Principal Component of a Rank-deficient Matrix
    M. Asteris, D. Papailiopoulos, G. N. Karystinos
    ISIT 2011. [IEEEXplore], [long], [slides]
  • Repairing Erasure Codes
    D. Papailiopoulos and A. G. Dimakis
    USENIX FAST 2011, Work-In-Progress (WiP). [short abstract]

2010

  • Distributed Storage Codes Meet Multiple-Access Wiretap Channels
    D. Papailiopoulos and A. G. Dimakis
    Allerton 2010. [IEEEXplore], [long], [slides]
  • MCMC Methods for Integer Least-Squares Problems
    B. Hassibi, A. G. Dimakis, and D. Papailiopoulos
    Allerton 2010. [IEEEXplore], [draft]
  • Interference Alignment as a Rank Constrained Rank Minimization
    D. Papailiopoulos and A. G. Dimakis
    GLOBECOM 2010. [IEEEXplore], [long], [slides], [MATLAB]
  • Maximum-likelihood Noncoherent OSTBC Detection with Polynomial Complexity
    D. Papailiopoulos and G. N. Karystinos,
    IEEE Transactions on Wireless Communications, June 2010. [IEEEXplore], [draft]

2007-2009

  • Optimal OSTBC Sequence Detection over Unknown Correlated Fading Channels
    D. Papailiopoulos and G. N. Karystinos
    Asilomar 2009. [IEEEXplore], [draft]
  • Efficient maximum-likelihood noncoherent orthogonal STBC detection
    D. Papailiopoulos and G. N. Karystinos
    Allerton 2008. [IEEEXplore], [draft], [slides]
  • Polynomial-complexity maximum-likelihood block noncoherent MPSK detection
    D. Papailiopoulos and G. N. Karystinos
    ICASSP 2008. [IEEEXplore], [draft]
  • Near ML detection of nonlinearly distorted OFDM signals
    D. Papailiopoulos and G. N. Karystinos
    Asilomar 2007. [IEEEXplore], [draft], [slides]

Invited Talks

  • Avoiding Coordination in Parallel Machine Learning
    Asilomar, 2016.
  • Let Your Cores Run Wild!: Avoiding Coordination in Parallel ML
    Stanford ISL Colloquium, 2016.
  • Less Talking, More Learning: Avoiding Coordination in Parallel ML
    ITA 2016.
  • Parallel Correlation Clustering on Big Graphs
    Allerton 2015.
  • Correlation Clustering on Big Graphs
    ITA 2015.
  • Big Graph Analytics through Low-rank Approximations
    ITA 2014 (graduation day).

Collaborators

I am very grateful to have collaborated with an amazing set of people, listed below in reverse chronological order.

Chris Ré (Stanford), Ce Zhang (ETH Zurich), Stephen Tu (UC Berkeley), Ramtin Pedarsani (UC Santa Barbara), Maximilian Lam (UC Berkeley), Kangwook Lee (KAIST), Siu On Chan (CUHK), Aviad Rubinstein (UC Berkeley) Horia Mania (UC Berkeley), Michael I. Jordan (UC Berkeley), Kannan Ramchandran (UC Berkeley), Ben Recht (UC Berkeley), Samet Oymak (Google), Xinghao Pan (UC Berkeley), Christos Boutsidis (Goldman Sachs), Anastasios Kyrillidis (UT Austin), Sriram Vishwanath (UT Austin), Ankit Singh Rawat (CMU), Constantine Caramanis (UT Austin), Ioannis Mitliagkas (Stanford), Dhruba Borthakur (facebook), Scott Chen (facebook), Ramkumar Vadali (facebook), Maheswaran Sathiamoorthy (Google), Stavros Korokythakis (Stochastic Technologies), Itzhak Tamo (Tel Aviv University), Guiseppe Caire (TU Berlin), Karthikeyan Shanmugam (UT Austin), Jin Li (MSR Redmond), Cheng Huang (MSR Redmond), Jianqiang Luo (Google), Georgina Abou-Elkheir (UPiraeus), Changho Suh (KAIST), Viveck Cadambe (Penn State), Asteris Megasthenis (UT Austin), Babak Hassibi (Caltech), Alex Dimakis (UT Austin), George Karystinos (TUC).


Contact

Address
1415 Engineering Drive, Engineering Hall,
University of Wisconsin-Madison
Madison, WI 53706
Email
dimitris@papail.io
dimitris@ece.wisc.edu