TensorNetwork

TensorNetwork copied to clipboard

Currently Tensor does not explicitly support the Symmetric backend. We should think about how to add such support. - There's an argument to be made that symmetric tensors differ significantly...

alewis

lgmres for numpy backend

3

mganahl

Jax eigsh_lanczos silently gives NaN when num_krylov_vecs > n

21

alewis

3- and N-body expectation values for MPS

3

Not sure how hard would this be, but it would be nice to have methods to calculate 3- and higher-order body expectation values in a matrix product state.

nquesada

mganahl

The MPS class should support sampling and computing logprobs (both finite and infinite). We should implement this at least for the numpy and JAX backends.

mganahl

benchmark eigs for JAX backend

2

We should check how eigs performs on CPU and GPU as compared to ARPACK

mganahl

Numpy decompositions tests

2

The tests in backends/numpy/decomposition_tests.py only test one dtype and pivot axis. Probably this should not be so.

alewis

mganahl

vuMPS

1

I'm now just one or two merges away from being able to start committing the example vuMPS code to the experimental_ncon branch, from which point on the latter will be...

alewis