Sparse data representation

Seminar zum Wissenschaftlichen Rechnen - Sommersemester 2011

Very often input and output data of a mathematical model, describing a physical process, are too much information. Imagine that the numerical solution (velocity, pressure, density etc.) of the given differential equation takes 10 MB of memory on your hard disk. If you repeat the numerical simulations a 10000 times (e.g. Monte Carlo simulations) then the required memory will be 100000 MB and that is huge. Thus, sparse data structures and sparse algorithms are required. One of the possible sparse data formats is a low-rank tensor format. Possible applications are in stochastic partial differential equations which model physical processes under uncertainties.

Kontakt