Skip to content

Bath compression #95

Description

@kbidzhiev

We can try compressing the baths with torch.eigh() or ~SVD.
Every bath element is a tensor with 3 legs $(\chi, mH_i, \chi)$ meaning 3d tensor, i.e. a list of Hermitian matrices $B(\chi, \chi)$ of length $mH_i$

These matrices can be decomposed with torch.eigh() $B_{\chi, \chi} = U S U^\dagger$, so that small eigenvalues(~singular values) can be chopped off, let's say 1e-10.

Again, this is a memory-performance trade. Maybe Multiplication $H_{eff} \psi$ will happen faster

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions