Some project and survey suggestions
Some project and survey suggestions
A few helpful links
You may find a webpage on
Geometry in Action
helpful.
These
geometry pages are a very good general resource, and include
links to bibliographies and search pages for them.
The
report
of the taskforce on application challenges for
computational geometry application may be helpful.
Potential survey topics
Surveys of the current literature for these areas might be of interest.
- geometric problems in statistics,
- including robust statistics: statistical estimators
that are not lead astray by a small number of input
values that are wildly wrong; possibly also spatial statistics.
- surface reconstruction:
- given a set of points in 3d, find a "reasonable" surface
containing those points
- kinetic computational geometry
- maintaining geometric information as input data moves
- external memory computational geometry
- handling geometric problems that have input data sets so large that
they don't fit in main memory.
- geometric problems in data mining and cluster analysis
- approximation of geometric objects
- geometric problems in approximating functions and in
scattered data interpolation
- geometric problems in geographic information systems
- visibility problems
- given a set of opaque objects and a viewpoint,
which can be seen? Can that information be updated
for a moving viewpoint?
- many-body simulation
Programming projects
For some of these problems, I have particular algorithms in mind;
for others, I can only point you to the literature for ideas.
For any of these, there are may be existing codes that might be adopted for use.
Java implementations of almost any of the algorithms described in class
would be of interest. Also:
- given a planar subdivision, or a polyhedron in 3d, subdivide it
into convex pieces; this is a hard problem, but the 2d
case is conceivable.
- "crusts" are a surface reconstruction technique (see above)
using Voronoi diagrams; implement them in 2d or 3d.
- given a convex polytope, approximate it with a polytope
having fewer vertices. Proposed algorithms do so within
a constant factor of best possible.
- given a set of points, compute their smallest enclosing sphere
or ellipsoid, or other smallest enclosing...
Known algorithms can do this in linear time.
- given two convex polytopes, compute the distance between them.
this can be done in linear time.
- given a set of points in d dimensions, compute the vertices
of their convex hull;
- implement and experiment with "natural neighbor interpolation"
methods, a technique for scattered data interpolation that uses
Voronoi diagrams.
- implement an algorithm for the "all-nearest-neighbors" problem:
given a set of sites (points), find, for every site, the nearest other site.
The algorithm uses quadtrees in a straightforward way.
Software Collections
The following collections of geometric software may be helpful or inspiring.
Ken Clarkson
August 1997