Fitting high-dimensional potential energy surface using active subspace and tensor train (AS+TT) method

Vitaly Baranov, Ivan Oseledets

    Research output: Contribution to journalArticlepeer-review

    5 Citations (Scopus)

    Abstract

    This paper is the first application of the tensor-train (TT) cross approximation procedure for potential energy surface fitting. In order to reduce the complexity, we combine the TT-approach with another technique recently introduced in the field of numerical analysis: an affine transformation of Cartesian coordinates into the active subspaces where the PES function has the most variability. The numerical experiments for the water molecule and for the nitrous acid molecule confirm the efficiency of this approach.

    Original languageEnglish
    Article number174107
    JournalJournal of Chemical Physics
    Volume143
    Issue number17
    DOIs
    Publication statusPublished - 7 Nov 2015

    Fingerprint

    Dive into the research topics of 'Fitting high-dimensional potential energy surface using active subspace and tensor train (AS+TT) method'. Together they form a unique fingerprint.

    Cite this