Dimitris Papailiopoulos

Postdoc @ UC Berkeley
AMPLab and BLISS

I will be joining the ECE Department at the University of Wisconsin-Madison as an Assistant Professor this Fall!
I am looking for graduate students to join my group!
If you are (or soon will be) a student at UW-Madison, and want to do research on the fascinating intersection of machine learning, parallel and distributed systems, and information theory, please email me!

About

I am a postdoc in the AMPLab and BLISS at UC Berkeley, working with Kannan Ramchandran and Ben Recht on large-scale machine learning problems. I received my Ph.D. in 2014 from UT Austin, and was fortunate to be advised by Alex Dimakis. Before UT, I spent 3.5 years as a grad student at USC.

I am interested in machine learning, coding theory, and parallel and distributed algorithms. My research focuses on large-scale data and graph analytics, coordination-free parallel ML algorithms and the use of codes to speed up distributed computation pipelines.


Curriculum Vitae (Jan. 2016)

News

May 2016
Preprint: 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 algorithms. 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.

Nov 2015
IEEE Signal Processing Society, 2015 Young Author Best Paper Award

for our paper with Alex Dimakis, Interference Alignment as a Rank Constrained Rank Minimization, appeared in IEEE Trans. on Signal Processing, Aug. 2012.

Publications

2016

  • 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 Trans. 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
    OPT Workshop, NIPS 2015. [long version]
  • Speeding-up Distributed Machine Learning Using Codes
    K. Lee, M. Lam, R. Pedarsani, Dimitris Papailiopoulos, Kannan Ramchandran
    ML-Systems Workshop, NIPS 2015. [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
  • 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
  • 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
  • Feedback in the K-user Interference channel
    D. Papailiopoulos, Changho Suh, Alexandros G. Dimakis
  • 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
  • Sparse Principal Component of a Rank-deficient Matrix
    M. Asteris, D. Papailiopoulos, G. N. Karystinos
  • Repairing Erasure Codes
    D. Papailiopoulos and A. G. Dimakis
    USENIX FAST 2011, Work-In-Progress (WiP) and Poster. [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 Trans. 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]

Teaching

Fall 2016

  • Stay Tuned!

Invited Talks

  • 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 have been very privileged to collaborate with the following great people, listed in chronological order of collaboration.

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


Contact

Address
465 Soda Hall (AMPLab), and 264 Cory Hall (WiFo),
University of California, Berkeley
Berkeley, CA 94709