studying networks through an algorithmic lens michael t. goodrich univ. of california, irvine
TRANSCRIPT
Studying Networks Studying Networks Through an Through an Algorithmic LensAlgorithmic Lens
Michael T. GoodrichUniv. of California, Irvine
A Paradigm ShiftA Paradigm Shift Transitioning from equations to algorithms as
the fundamental model for prediction
Images from wikimedia.org and computerhistory.org
2000
E=mc2
Equations
Algorithms
F=ma
19451700
The Algorithmic LensThe Algorithmic Lens
Image from http://uwnews.org/uweek/uweekarticle.asp?visitsource=uwkmail&articleID=39100
Exposes the computational nature of natural processes and provides a language for their description.
Brings to bear fundamental algorithmic concepts: adversarial and probabilistic models, asymptotic analysis, intractability, computational learning theory, threshold behavior.
Provides a new worldview for many fields.
Two Flavors of the Algorithmic LensTwo Flavors of the Algorithmic Lens
1. Discover inherent combinatoric, topological and geometric properties of various networks that can improve algorithms that operate on such networks.
2. The algorithmic worldview provides new computational insights, models, and metaphors about networks
Image by Argus fin from http://commons.wikimedia.org/wiki/Image:International_E_Road_Network.png, and is in the public domain
Social Sciences and NetworksSocial Sciences and Networks The Web is a powerful laboratory for studying social and
economic systems as computational processes. Insights from algorithmic theory are indispensable for
understanding interactions and relationships that the Internet has spawned.
Images from wikimedia.org, in public domain,under CC license and GNU FDL
Kevin BaconBarack Obama
The World is Not Flat (or Spherical)The World is Not Flat (or Spherical) Road networks are built on the surface of the Earth, which
is like a sphere. Previous researchers have viewed road networks as plane graphs, that is, graphs that are drawn on a sphere without edge crossings. (E.g., see [Chou, 96].)
Image by NASA, is available at http://commons.wikimedia.org/wiki/Image:Europe_terrain.jpg, and is in the public domain
The World is Not Flat (or Spherical)The World is Not Flat (or Spherical) Unfortunately, real road networks are highly non-planar. In particular, a road network with n vertices typically has a
number of edge crossing proportional to
)( nO
Data is from the U.S. TIGER/Line road network database, as provided by the Ninth DIMACS Implementation Challenge
The Geometry of Road NetworksThe Geometry of Road Networks So we want a new geometric way to characterize road
networks that leads to improved algorithms Previous researchers have made assumptions about edge weights,
such as their being Euclidean or having nice distributions (e.g., [Goldberg, 93], [Holzer et al., 05], [Klunder and Post, 06], [Sanders and Schultes, 05], [Sedgewick and Vitter, 86], [Thorup, 99]).
Instead, we took a different approach inspired by the work of [Eppstein et al., 93] and [Miller et al., 95, 97, 98] on separators in disk neighborhood systems.
Image by Pbroks13, from http://commons.wikimedia.org/wiki/Image:Apollonian_circles.svg, used under the GFDL 1.2
Disk Neighborhood SystemsDisk Neighborhood Systems Given a collection of disks in the plane, define a graph G
so that each disk is associated with a vertex in G and there is and edge between v and w if their disks intersect.
Image by David Eppstein, available at http://commons.wikimedia.org/wiki/Image:Unit_disk_graph.svg, in the public domain
For each vertex v, define a disk with radius equal to half the length of the longest road adjacent to v.
The road network is guaranteed to be a subgraph of this Natural Disk Neighborhood System.
The Natural Disk Neighborhood The Natural Disk Neighborhood System for Road NetworksSystem for Road Networks
Data is from the U.S. TIGER/Line road network database, as provided by the Ninth DIMACS Implementation Challenge
Future DirectionsFuture Directions Study additional algorithmic properties of networks,
combining combinatorics, geometry, and topology. Use the algorithm as a part of the statistical models
for social interactions