mps_common
full name: tenpy.algorithms.mps_common
parent module:
tenpy.algorithms
type: module
Classes
|
Mixer based on reduced density matrices. |
|
A dummy replacement for |
|
Prototype class for local effective Hamiltonians used in sweep algorithms. |
|
Prototype class for algorithms that iterate the same sweep until convergence. |
|
Base class for a general Mixer. |
|
Class defining the one-site effective Hamiltonian for Lanczos. |
|
Variational MPO application, using QR-based decompositions instead of SVD. |
|
Mixer of a direct subspace expansion. |
|
Prototype class for a 'sweeping' algorithm. |
|
Class defining the two-site effective Hamiltonian for Lanczos. |
|
Variational compression for applying an MPO to an MPS (in place). |
|
Variational compression of an MPS (in place). |
|
Class defining the zero-site effective Hamiltonian for Lanczos. |
Module description
‘Sweep’ algorithm and effective Hamiltonians.
Many MPS-based algorithms use a ‘sweep’ structure, wherein local updates are performed on the MPS tensors sequentially, first from left to right, then from right to left. This procedure is common to DMRG, TDVP, MPO-based time evolution, etc.
Another common feature of these algorithms is the use of an effective local Hamiltonian to perform the local updates. The most prominent example of this is probably DMRG, where the local MPS object is optimized with respect to the rest of the MPS-MPO-MPS network, the latter forming the effective Hamiltonian.
The Sweep
class attempts to generalize as many aspects of ‘sweeping’
algorithms as possible. EffectiveH
and its subclasses implement the
effective Hamiltonians mentioned above. Currently, effective Hamiltonians for
1-site and 2-site optimization are implemented.
The VariationalCompression
and VariationalApplyMPO
implemented here also directly use the Sweep
class.