Bio


Mohsen Bayati received his PhD in Electrical Engineering from Stanford University in 2007. His dissertation was on machine learning and modeling aspects of large-scale networks. During the summers of 2005 and 2006 he interned at IBM Research and Microsoft Research respectively.

He was a Postdoctoral Researcher with Microsoft Research from 2007 to 2009 working mainly on applications of machine learning and optimization methods in healthcare. In particular, he focused on hospital readmissions. Nearly one in every five patients is readmitted to the hospital within 30 days of their discharge. The estimated cost of unplanned rehospitalizations to Medicare in 2004 was around $17.4 billion. Mohsen Bayati and his colleagues at Microsoft Research applied machine learning methods to hundreds of thousands of hospital electronic health records to identify patients with the highest risk of being rehospitalized and obtained a decision support mechanism for allocating scarce resources to post-discharge support.

He has been a Postdoctoral Scholar at Stanford University from 2009 to 2011 with a research focus in high-dimensional statistical models.

Academic Appointments


  • Assistant Professor, Operations, Information & Technology
  • Assistant Professor (By courtesy), Electrical Engineering

Administrative Appointments


  • Assistant Professor, Stanford Univerisity (2011 - Present)
  • Postdoctoral Scholar, Stanford University (2009 - 2011)
  • Postdoctoral Researcher, Microsoft Research (2007 - 2009)
  • Research Intern, Microsoft Research (2006 - 2006)
  • Research Intern, IBM Research (2005 - 2005)

Honors & Awards


  • MBA Class of '69 Faculty Scholar, Stanford Graduate School of Business (2012-13)
  • Gold Medal, International Mathematics Olympiad (1997)

Current Research and Scholarly Interests


For a full description of some of my current projects, please visit my homepage at:

http://www.stanford.edu/~bayati/

Projects


  • Data-driven decision making in healthcare, Stanford University/Microsoft Research (3/1/2009 - Present)

    Rehospitalization -patient admission to a hospital soon after the discharge- is both common and costly. Nearly one in every five patients is readmitted to the hospital within 30 days of their discharge. The estimated cost of unplanned rehospitalizations to Medicare in 2004 was around $17.4 billion. Research shows that hospital initiatives such as: patient education programs, follow-up home visits by pharmacists, extensive discharge packages, etc., can avert many rehospitalizations. However, proper allocation of these costly and limited resources is a challenging problem. In this project, we use machine learning and optimization tools for identifying patients with highest risk of being rehospitalized and obtain a decision support mechanism for allocating scarce resources to post-discharge support.

    Location

    Stanford, CA and Cambridge, MA

  • Modeling of large networks with given constraints, Stanford University (10/1/2005 - 9/30/2009)

    A central hurdle in working with large networks is the lack of good random graph models that capture their specific properties. Such random graph models can be useful for example in simulating algorithms for the Internet or detecting motifs (patterns) in biological systems. Unfortunately, the existing algorithms for generating random graphs with given degrees have large running times (quadratic on problem size), making them impractical to use for networks with millions of nodes. In this paper we designed the fastest (almost linear running time) algorithm and proved its correctness when the node degrees are in certain range. Recently, with Andrea Montanari and Amin Saberi, we extended that algorithm to design an efficient algorithm for generating random graphs with no short cycles that appeared in SODA 2009. These graphs are used to construct high performance codes that can achieve the Shannon capacity.

    Location

    Stanford, CA

  • Fast algorithms for aligning large networks, Stanford University/Microsoft Research (1/1/2008 - 10/30/2011)

    Network alignment generalizes and unifies several approaches for forming a matching or alignment between the nodes of two networks. We propose a mathematical programming framework for network alignment problem and a sparse variation of it where only a small number of matches between the nodes of the two networks are possible. We propose new message passing algorithms that allow us to compute, very efficiently, near optimal solutions to the sparse network alignment problems with network sizes as large as hundreds of thousands of nodes. Using (rigorous) linear programming upper-bounds we show that our algorithms produce near optimal solutions on problems in bioinformatics, in ontology matching, and on synthetic problems.

    Location

    Stanford, CA and Cambridge, MA

2014-15 Courses


Journal Articles


  • Message-Passing Algorithms for Sparse Network Alignment ACM TRANSACTIONS ON KNOWLEDGE DISCOVERY FROM DATA Bayati, M., Gleich, D. F., Saberi, A., Wang, Y. 2013; 7 (1)
  • The LASSO Risk for Gaussian Matrices IEEE TRANSACTIONS ON INFORMATION THEORY Bayati, M., Montanari, A. 2012; 58 (4): 1997-2017
  • Predictive Models and Policies for Minimizing Rehospitalizations for Congestive Heart Failure Working paper Bayati, M., Braverman, M., Gillam, M., Smith, M., Horvtiz, E., et al 2012
  • BELIEF PROPAGATION FOR WEIGHTED b-MATCHINGS ON ARBITRARY GRAPHS AND ITS RELATION TO LINEAR PROGRAMS WITH INTEGER SOLUTIONS SIAM JOURNAL ON DISCRETE MATHEMATICS Bayati, M., Borgs, C., Chayes, J., Zecchina, R. 2011; 25 (2): 989-1011

    View details for DOI 10.1137/090753115

    View details for Web of Science ID 000292302000033

  • A Sequential Algorithm for Generating Random Graphs ALGORITHMICA Bayati, M., Kim, J. H., Saberi, A. 2010; 58 (4): 860-910

Conference Proceedings