AT&T Labs Research

Yifan Hu received a BSc/MSc in Applied Mathematics from Shanghai Jiao-Tong University in 1985/1988. After a brief period of teaching at Shanghai Jiao-Tong, he went on to receive a Ph.D. in Optimization from Loughborough University in 1992. He worked as a senior scientific officer in Daresbury Laboratory between 1992 to 2001, with a short period working for Computational Dynamics in between. He joined Wolfram Research in 2001, and AT&T Labs in 2007. At present, he is a Principal Member of Technical Staff. His research interests include network/information visualization, machine learning, numerical and combinatorical algorithms and scientific computing. Further information can be found at his personal website.

Projects
Graphviz System for Network Visualization, Graphviz is network visualization software for creating high-quality, readable node-link diagrams of large-scale data sets

Information Visualization Research - Prototypes and Systems, describes work on innovative prototypes and systems

This version of the work is reprinted here with permission of IEEE for your personal use. Not for redistribution. The definitive version was published in 2012 , 2012-09-30

{For embedding graphs or high dimensional data, stress model is widely used when edges length, or distances between items, are specified. However, traditional full stress model is not scalable, due to the need for all-pairs shortest path calculation. A number of fast approximation algorithms were proposed. While they work well for some graphs, on graphs of intrinsic high dimensions, such as some non-rigid graphs, the results are less satisfactory. In this paper we propose a maxent-stress model. The method uses the principal of maximal entropy to deal with the extra degrees of freedom. We formulate a force-augmented stress majorization algorithm to solve the maxent-stress model. Numerical results show that the algorithm can scale to large graphs, yet does not degrade on non-rigid graphs. This also has potential applications to scalable algorithms for statistical multidimensional scaling (MDS) with variable distances. }

This version of the work is reprinted here with permission of IEEE for your personal use. Not for redistribution. The definitive version was published in IEEE Transactions on Visualization and Computer. , 2011-07-01

{Maps offer a familiar way to present geographic data (continents, countries, states), and additional information (topography, geology, rainfall) can be displayed with the help of contours and heat-map overlays. In this paper we consider visualizing large-scale dynamic relational data by taking advantage of the geographic map metaphor. We describe a system that visualizes user traffic on the Internet radio station last.fm and address challenges in mental map preservation, as well as issues in animated map-based visualization.}

This version of the work is reprinted here with permission of IEEE for your personal use. Not for redistribution. The definitive version was published in IEEE Transactions on Visualization and Computer Graphics. , 2011-07-01, http://www.computer.org/portal/web/tvcg/

{Maps offer a familiar way to present geographic data (continents, countries), and additional information (topography, geology), can be displayed with the help of contours and heat-map overlays. In this paper we consider visualizing large-scale dynamic relational data by taking advantage of the geographic map metaphor. We describe a map-based visualization system which uses animation to convey dynamics in large datasets, and which aims to preseve the viewer's mental map while also offering readable views at all times. Our visualization system is fully functional and has been used to visualize user traffic on the Internet radio station last.fm, as well as TV-viewing patterns from an IPTV service. All map images in this paper are available in high-resolution at http://www2.research.att.com/~yifanhu/TrendMap as are several movies illustrating }

(c) ACM, 2011. This is the author's version of the work. It is posted here by permission of ACM for your personal use. Not for redistribution. The definitive version was published in ACM Transactions on Mathematical Software , 2011-01-01.

{We describe the University of Florida Sparse Matrix Collection, a large and actively growing set of sparse matrices that arise in real applications. The Collection is widely used by the numerical linear algebra community for the development and performance evaluation of sparse matrix algorithms. It allows for robust and repeatable experiments: robust because performance results with artificially-generated matrices can be misleading, and repeatable because matrices are curated and made publicly available in many formats. Its matrices cover a wide spectrum of domains, include those arising from problems with underlying 2D or 3D geometry (as structural engineering, computational fluid dynamics, model reduction, electromagnetics, semiconductor devices, thermodynamics, materials, acoustics, computer graphics/vision, robotics/kinematics, and other discretizations) and those that typically do not have such geometry (optimization, circuit simulation, economic and financial modeling, theoretical and quantum chemistry, chemical process simulation, mathematics and statistics, power networks, and other networks and graphs). We provide software for accessing and managing the Collection, from MATLAB, Mathematica, Fortran, and C, as well as an online search capability. Graph visualization of the matrices is provided, and a new multilevel coarsening scheme is proposed to facilitate this task. }

{Graphs are often used to encapsulate relationships between objects. Node-link diagrams, commonly used to visualize graphs, suffer from visual clutter on large graphs. Edge bundling is an effective technique for alleviating clutter and revealing high-level edge patterns. Previous methods for general graph layouts either require a control mesh to guide the bundling process, which can introduce high variation in curvature along the bundles, or all-to-all force and compatibility calculations, which is not scalable. We propose a multilevel agglomerative edge bundling method based on a principled approach of minimizing ink needed to represent edges, with additional constraints on the curvature of the resulting splines. The proposed method is much faster than previous ones, able to bundle hundreds of thousands of edges in seconds, and one million edges in a few minutes. }

This version of the work is reprinted here with permission of IEEE for your personal use. Not for redistribution. The definitive version was published in 5th IEEE Pacific Visualization Symposium. , 2011-12-01

The definitive version was published in 19th International Symposium on Graph Drawing. , 2011-12-01

{We describe a practical approach for visualizing dynamic relational data and multiple relationships on the same data using a geographic map metaphor, where clusters of nodes form countries and neighboring countries correspond to nearby clusters. Our aim is to compare two or more maps obtained using differ- ent similarity metrics and to provide an interactive tool to visually explore the effect of combining two or more similarity metrics. Our method ensures good readability and mental map preservation, based on dynamic node placement with node stability, dynamic clustering with cluster stability, and dynamic coloring with color stability. }

(c) ACM, 2011. This is the author's version of the work. It is posted here by permission of ACM for your personal use. Not for redistribution. The definitive version was published in KDD CUP 2011 workshop , 2011-08-21.

{We describe our solution for the KDD Cup 2011 track 2 challenge. Our solution relies heavily on ensembling together diverse individual models for the prediction task, and achieved a final leaderboard misclassification rate of 3.8863\%. This paper provides details on both the modeling and ensemble creation steps.}

The definitive version was published in 18th International Symposium on Graph Drawing , Volume 6502, 2010-10-22

{In this paper, we consider the problem of representing graphs by triangles whose sides touch. We present linear time algorithms for creating touching triangle representations for outerplanar graphs, square grid graphs, and hexagonal grid graphs. The class of graphs with touching triangle representations is not closed under minors, making characterization difficult. We do show that pairs of vertices can only have a small common neighborhood, and we present a complete characterization of the subclass of biconnected graphs that can be represented as triangulations of some polygon.}

AT&T Labs Research — Leading Invention, Driving Innovation

site map