Variational MPO compression for optimizing 2-D isometric tensor networks.

ORAL

Abstract

A scheme for ground-state optimization of two-dimensional isometric tensor networks. The effective Hamiltonian for each column is treated as a Matrix Product Operator, and truncated using a variational MPO compression algorithm which "weights" the optimization by the reduced density matrix of the next column, enabling a stronger reduction of the MPO bond dimension.

–

Presenters

  • Gabriel Woolls

    • University of California, Berkeley

Authors

  • Gabriel Woolls

    • University of California, Berkeley

  • Zhehao Dai

    • University of Pittsburgh

  • Sheng-Hsuan Lin

    • TU Munich

  • Yantao Wu

    • University of California, Berkeley
    • Chinese Academy of Sciences
    • Chinese Academy of Science

  • Michael P Zaletel

    • University of California, Berkeley