Carnegie Mellon University
Computer Science
SELECT Lab

Home

 

 

 

 

 

New! Join our LinkedIn Group

.

 

Theory Papers

  • Gaussian Belief Propagation: Theory and Application. D. Bickson. Ph.D. Thesis. Submitted to the senate of the Hebrew University of Jerusalem, October 2008. Revised July 2009. arxiv
  • Fault identifiction via non-parametric belief propagation. D. Bickson, D. Baron, A. Ihler, H. Avissar and D. Dolev. IEEE Tran. on Signal Processing arxiv
  • Distributed Sensor Selection using a Truncated Newton Method. D. Bickson and D. Dolev. arxiv
  • A low density lattice decoding via non-parametric belief propagation. D. Bickson, A. Ihler and D. Dolev. 47h Annual Allerton Conference on Communication, Control and Computing, Allerton House, Illinois, Sept. 2009, pp. 439-446. arxiv bibtex
  • Fixing onvergence of Gaussian belief propagation algorithm. J. K. Johnson, D. Bickson and D. Dolev. Proc. of the International symposium on information theory (ISIT), Vol. 3, June 28 - July 3, 2009, Coex, Seoul, Korea, pp. 1674-1678. arxiv bibtex
  • Distributed large scale network utility maximization. D. Bickson, Y. Tock, A. Zymnis, S. Boyd and D. Dolev. Proc. of the International symposium on information theory (ISIT), Vol. 2, June 28 - July 3, 2009, Coex, Seoul, Korea, pp. 829-833. arxiv bibtex
  • Gaussian belief propagation for solving systems of linear equations: theory and application. O. Shental, D. Bickson, P.H. Siegel, J.K. Wolf and D. Dolev. Tech Report. arxiv bibtex
  • Distributed Kalman filter via Gaussian belief propagation. D. Bickson, O. Shental and D. Dolev, In the 46th Annual Allerton Conference on Communication, Control and Computing, Allerton House, Illinois, Sept. 2008, pp. 628-635. arxiv bibtex
  • Polynomial linear programming with Gaussian belief propagation. D. Bickson, Y. Tock, O. Shental and D. Dolev, In the 46th Annual Allerton Conference on Communication, Control and Computing, Allerton House, Illinois, Sept. 2008, pp. 895-901. arxiv bibtex
  • Gaussian belief propagation solver for systems of linear equations. O. Shental, D. Bickson, P. H. Siegel, J. K. Wolf, and D. Dolev, In IEEE Int. Symp. on Inform. Theory (ISIT), Toronto, Canada, July 2008, pp. 1863 - 1867. arxiv bibtex
  • Gaussian belief propagation based multiuser detection. D. Bickson, O. Shental, P. H. Siegel, J. K. Wolf, and D. Dolev, In IEEE Int. Symp. on Inform. Theory (ISIT), Toronto, Canada, July 2008, pp. 1878-1882. arxiv bibtex
  • Linear Detection via Belief Propagation. Danny Bickson, Danny Dolev, Ori Shental, Paul H. Siegel and Jack K. Wolf. In the 45th Annual Allerton Conference on Communication, Control, and Computing, Allerton House, Illinois, Sept. 07, pp. 1207-1213. pdf bibtex
  • A message-passing solver for linear systems, O. Shental, D. Bickson, P. H. Siegel, J. K. Wolf, and D. Dolev, In Proc. Information Theory and Applications (ITA) Workshop, San Diego, CA, USA, January 2008. pdf bibtex

    Large-Scale Applications

  • GraphLab: A New Parallel Framework for Machine Learning. Yucheng Low, Joseph Gonzalez, Aapo Kyrola, Danny Bickson, Carlos Guestrin, and Joseph M. Hellerstein (2010). In the 26th Conference on Uncertainty in Artificial Intelligence (UAI), 2010. pdf bibtex
  • A statistical approach to monitoring of soft-real time distributed systems. D. Bickson, G. Gershinsky, E. Hoch and K. Shagin. In ICDCS 2009, submitted for publication. Nov. 2008. arxiv bibtex
  • A unifying framework for rating users and data items in Peer-to-Peer and social networks. Danny Bickson and Dahlia Malkhi. In Peer-to-Peer Networking and Applications (PPNA) Journal, Accepted, January 2008. html bibtex
  • Peer–to-Peer Rating. Danny Bickson, Dahlia Malkhi and Lidong Zhou. In the 7th IEEE Peer-to-Peer Computing, Galway, Ireland, Sept. 2007. pdf bibtex
  • A Gaussian Belief Propagation Solver for Large Scale Support Vector Machines. D. Bickson, D. Dolev and E. Yom-Tov. In the 5th European Complex Systems Conference.,Sept. 2008. pdf bibtex

    Other Resources Presentation given by Prof. Erik Aurell from KTH (Sweden) in Workshop on Techniques and Challenges from Statistical Physics CRM@UAB, Barcelona, Obtober 2009. pdf

    Source Code

    LARGE SCALE/MULTICORE GAUSSIAN BP C++ CODE IS AVAILABLE as part of the GraphLab project: here.
    Are you using my code? Joing our LinkedIn group

  • Gaussian belief propagation Matlab package
    gabp-src.zip
    This package was downloaded aleardy more than 1,000 times! If you are using my code I will be happy to hear what are you working on and provide limited support for the projects I find interesting.

    Written by Danny Bickson.

    If you are using my code, please cite the following: @phdthesis{bicksonThesis title={Gaussian Belief Propagation: Theory and Application}, author={Bickson, D.}, year={2008}, school={The Hebrew University of Jerusalem}, address = {Jerusalem, Israel}, }

    The following algorithms are implemented:
    gabp.m, run_gabp.mGaussian BP - parallel version
    asynch_GBP.mGaussian BP - serial version
    sparse_gabp.m, run_sparse_gabp.mGaussian BP - sparse version, optimized, tested on sparse matrices of size 0.5M x 0.5M , with 4% non zeros
    gabpms.ms, run_gabpms.mQuadratic Min-Sum algorithm - Moallemi and Van-Roy
    LP/Linear programming using GaBP (Allerton 2008 paper) arxiv
    GaBP_convergence_fix/Fixing convergence of GaBP algorithm, mulituser detection example (ISIT 2009) arxiv
    LDLC/Low density lattice decoder (LDLC) using gaussian mixtures
    NUM/Distributd large scale network utility maximization (ISIT 2009) arxiv
    ILP/Non-parametric belief propagation (NBP) implementation via Alex Ihler's Matlab KDE toolbox.
    NBP/Non-parametric belief propagation (NBP) implementation via quantization (more efficient), including working compressive sensing example and boolean least squares (multiuser detection) example. This code was extensively tested in Dror Baron's compressive sensing journal paper
    SS/Distributed Sensor Selection using a Truncated Newton Method. Submitted for publication. arxiv
    NBP_decoder/Non-parametric belief propagation (NBP) decoder for low density lattice codes. Allerton 2009. arxiv
    fault_detection/Distributed fault detection via NBP. paper website
    KRR/Kernel ridge regression (for fitting a curve to a Gaussian mixture, needed as input to NBP)
    Lanczos/Matlab implementation of the Lanczos algorithm for computing eigenvalues
    bptf_demo/Matlab implementation of alternating least squares matrix factorization

    Acknowledgements
  • This research was partially supported by ISF (Israeli Science Foundation) grant number 0397373. NSF IIS-0803333, NSF NeTS-NBD CNS-0721591 and DARPA IPTO FA8750-09-1-0141
  • LDLC/ILP code relies on Alex Ihler's Matlab KDE package found here
  • The LDLC algorithm was implemented with the great help of Naftali Sommer, Tel Aviv University.
  • NBP/LDLC encoding matrices were kindly provided by Marilynn Green, Nokia Siemens Networks Research Technology Platforms Dallas, TX.
  • The non-parametric BP implementation heavily relies on Dror Baron's compressive sensing decoder Further information is found on the fault identification with compressed sening page.
  • The NUM simulation and the fault detection simulation was originally written by Argyris Zymnis - found here
  • An excellent discrete BP matlab code package by Talya Meltzer is found here
  • The sensor selection simulation was originally written by Joshi - found here

    Are you using my code? Drop me a note! I would like to hear about it! Danny DOT Bickson @ GMAIL DOT COM

    Number of unique visitors since April 19, 2009: HTML Hit Counter
    Web Counters

    • Original site design by Pegasus Web Design Resources, modified by David Rabinowitz