A Comparison of Adaptive Refinement Schemes for Numerical PDE
Solvers
David Lenz
UCSD
Abstract:
PDEs are often solved numerically by making a guess for the
solution and then continually modifying that guess so as to better
approximate the true solution. This process of modification usually
involves changing the function space from which the approximation is
drawn. There are many ways that the approximation space could be
changed to best reduce error, and one may even wish to utilize more
than one in the same problem. In this talk I will describe three
strategies for refining approximation spaces and compare their
performance on problems of different types.