On the 14th of January, Michael Götte gave a talk about Xerus, a C++ tensor library. The library is able to perform numerical computations with higher order tensors, as well as Tensor-Train Decomposition and is also useful for general tensor networks. Current usage of the library includes uncertainty quantification, electronic structure calculations and machine learning, as well as fluid dynamics. The talk started with an introduction to tensors, the tensor product and tensor contraction, and outlined mathematical examples. The mathematical examples were then computed using the Xerus library, giving an overview over the basic syntax and usage of the library, using Python as a frontend. After the basics were covered, more advanced concepts were discussed, such as tensor trains and tensor hierarchies. The theoretical discussion was again supplemented with numerical examples, introducing the Xerus syntax to compute Tensor-Train Decompositions and the creation of graphical representations for tensor hierarchies.
Tools Seminar: Xerus – A C++ Tensor Library