Videos from JuliaCon are now available online
2018
Previous editions: 2017 | 2016 | 2015 | 2014
Frank Otto

University College London



Hierarchical Tensor Decompositions in Julia

Tensor decompositions are a powerful method to compactly represent high-dimensional data. Such data is encountered quite naturally in the numerical treatment of quantum systems with many particles, e.g. small molecules. In this talk I will show how Julia helped me to represent a nine-dimensional potential energy surface for the H3O2- molecule as a compact hierarchical tensor, which allowed for a very efficient numerical computation of its ground state. The resulting Julia code for computing hierarchical tensor decompositions may be of general use for dealing with structured high-dimensional data.

Speaker's bio

Frank has a PhD in Theoretical Chemistry, and many years of experience with scientific software development and high performance computing. In 2017 he joined the Department of Chemistry at University College London as an HPC support specialist. Frank has been keenly following the development of Julia since 2013, and continues to advocate its use to his colleagues.