NDTensors.jl

NDTensors.jl copied to clipboard

There are some more contraction logic optimizations to do, following up on #25. For example: - [ ] Have `ContractionProperties` only store Tuples, not MVectors. MVectors can be used as...

mtfishman

Try to simplify the overload requirements for inds

4

We should reevaluate the interface requirements for the `inds` object of the `Tensor` type, and try to make it as simple as possible. For example, if we assume the `inds`...

mtfishman

It might be nice to have a lazily permuted Tensor wrapper, similar to Julia's `PermutedDimsArray`. We could define it as something like `PermutedDims{

mtfishman

Just a a reminder to the discussion in https://github.com/ITensor/NDTensors.jl/pull/14. I could do it at some point this week.

orialb

Right now, `U,S,V = svd(A)` has the convention `U*S*transpose(V) == A`, while Julia uses the convention that it outputs `V` such that `U*S*V' == A`. We should consider changing it...

mtfishman

Think about the names of sparse storage types/tensors

1

It may be good to rethink the names of the sparse storage tensors and constructors. Right now there is: - `Dense` - `Diag` - `BlockSparse` - `DiagBlockSparse` `DiagBlockSparse` may not...

mtfishman

Currently, `permutedims!!(R::Tensor, T::Tensor, perm, f::Function)` accepts only a 2-argument function that applies pairwise to the elements of `R` and `permutedims(T, perm)`. This could be replaced by a more general function...

mtfishman

mtfishman