Tensor backend
This internal component implements the tensors and contractions used by SpinGlassPEPS.jl.
We don't expect the user to interact with this package, as it is more of a "back-end" type. Nevertheless, we provide API references should the need arise.
The tensor backend:
- includes the creation and manipulation of various types of tensors.
- offers efficient tools for tensor contractions, in particular it handles tensor contractions in a sparse and dense mode, enabling efficient memory and computational use depending on the problem's structure.
- supports QMps (Matrix Product States) and QMpo (Matrix Product Operators) for representing states and operators in a tensor network framework.
- implements advanced tensor optimization algorithms, including zipper algorithm for approximating boundary MPS and variational schemes.
- supports computation on both CPU and GPU, enabling high-performance tensor operations with the
onGPUflag in core tensor structures.