Nate Veldt


I am a sixth year Mathematics PhD student at Purdue University. My advisor is Professor David F. Gleich of Purdue's computer science department.

My research interests lie in network analysis, optimization, and machine learning. In particular I've done a lot of work recently on algorithms for graph clustering.

In the fall of this year I will begin a postdoctoral position in the Center for Applied Mathematics at Cornell University, working under the supervision of Professors Austin Benson and Jon Kleinberg.

Curriculum vitae

Email: lveldt--at--purdue--dot--edu

Githubhttps://github.com/nveldt

Twitterhttps://twitter.com/n_veldt

 

     Publications

    1. Metric-Constrained Optimization for Graph Clustering Algorithms
      Nate Veldt, David Gleich, Anthony Wirth and James Saunderson
      SIAM Journal on Mathematics of Data Science (accepted for publication, 2019)
      arXiv Preprint
    2. Learning Resolution Parameters for Graph Clustering
      Nate Veldt, David Gleich, and Anthony Wirth
      Proceedings of the 2019 World Wide Web Conference (to appear May 2019)
      preprint
    3. Flow-Based Local Graph Clustering with Better Seed Set Inclusion
      Nate Veldt, Christine Klymko, and David Gleich
      Proceedings of the 2019 SIAM International Conference on Data Mining (to appear May 2019)
      preprint
    4. Correlation Clustering Generalized
      David Gleich, Nate Veldt, and Anthony Wirth
      Proceedings of the 29th International Symposium on Algorithms and Computation, December 2018,
      paper
    5. A Correlation Clustering Framework for Community Detection
      Nate Veldt, David Gleich, and Anthony Wirth
      Proceedings of the 27th International World Wide Web Conference, April 2018
      paper
    6. full version
    7. Low-Rank Spectral Network Alignment
      Huda Nassar, Nate Veldt, Shahin Mohammadi, Ananth Grama, David Gleich
      Proceedings of the 27th International World Wide Web Conference, April 2018
      paper
    8. Correlation Clustering with Low-Rank Matrices
      Nate Veldt, Anthony Wirth, and David Gleich
      Proceedings of the 26th International World Wide Web Conference, April 2017
      paper
    9. full version
    10. A Simple and Strongly Local Flow-Based Method for Cut Improvement
      Nate Veldt, David Gleich, and Michael Mahoney
      Proceedings of the 33rd Annual International Conference on Machine Learning, June 2016
      paper Conference Talk

     Preprints

    1. A Parallel Projection Method for Metric-Constrained Optimization
      Cameron Ruggles, Nate Veldt, and David Gleich
      arXiv Preprint

    Recent and Upcoming Events

    I'll be traveling to a few different conferences this spring and summer to present my research. Other News:

    Research

    My research is focused on algorithms and complexity results for problems arising in network analysis and data science. This broadly includes work in mathematical optimization, machine learning, matrix computations, and theoretical computer science. To date much of my research has involved the theory and application of clustering algorithms. I have specifically worked on special variants of correlation clustering for partitioning signed datasets, flow-based methods for localized community detection, and fast solvers for convex relaxations of graph clustering objectives. My motivation is to bridge the gap between the best theoretical results and the most practical algorithms for problems in network science and data mining. An overarching goal is to develop methods that are fast, satisfy strong approximation guarantees, and explicitly take into account important features of the real-world networks and datasets they operate on.

    Teaching

    I am not teaching this semester, but here is a list of my previous teaching experience.

    In Fall of 2015 I received the Purdue Mathematics Department Excellence in Teaching Award:

    Purdue Math Department Excellence in Teaching Award 2015

    Purdue University Numerical Linear Algebra Group

    From Fall 2016 to Spring 2017 I organized and ran the Purdue University Numerical Linear Algebra Group, a seminar series on matrix computations and network analysis. All PUNLAG talks are recorded and uploaded to YouTube. The seminar webpage is here. You can subscribe to the Purdue Numerical Linear Algebra YouTube channel at PurdueNLA.

     

    Recorded Seminar Talks

    I frequently give local seminar talks on topics in network analysis and numerical linear algebra (often as a part of the PUNLAG seminar mentioned above). Sometimes the talks are on a specific research project I'm working on. Often though I like to take a tool or technique that has been useful in my research and present it in a way that can be more broadly applied to other projects and problems that people might be interested in. Here are some recorded talks on optimization techniques that have been useful in my work.

    Totally Unimodular Matrices in Linear Programming

    Solving Low-Dimensional Optimization Problems via Zonotope Vertex Enumeration

    Efficiently Solving Linear Programs with Triangle Inequality Constaints