Tensor product decomposition methods for noise reduction, data compression, multi-scale analysis, and projective integration

Diego del-Castillo-Negrete
Oak Ridge National Laboratory

Tensor product decomposition (TPD) methods are a powerful linear algebra technique for the efficient representation of high dimensional data sets. In the simplest 2-D case, TPD reduces to the well-known singular value decomposition of matrices. We discuss the application of TPD methods to: (i) Noise Reduction in particle-based computations; (ii) Data Compression for 3-D Magnetohydrodynamic and 6-D Gyrokinetic simulations; (iii) Multiscale Analysis of plasma turbulence; (iv) Projective Integration in collisional transport computations.


Back to Workshop II: Computational Challenges in Magnetized Plasma