adding permute function for reordering tensor products of quantum objects #151
Conversation
|
Sorry that we have re-organized the source code structure yesterday. So there's a conflict compare to the main branch. |
src/quantum_object.jl
Outdated
| ``` | ||
|
|
||
| """ | ||
| function permute end |
There was a problem hiding this comment.
Maybe just directly put the definition of permute here, we can merge those three methods in just one line, namely
function permute(
A::QuantumObject{<:AbstractArray{T},ObjType},
order::AbstractVector{Int}
) where {T,ObjType<:Union{KetQuantumObject,BraQuantumObject,OperatorQuantumObject}}
if length(order) != length(A.dims)
throw(ArgumentError("The order vector must have the same length as the number of subsystems"))
end
perm = kron_permutation(order, A.dims)
return QuantumObject(A.data[perm, perm], A.type, A.dims[order])
end
src/quantum_object.jl
Outdated
| # helper functions | ||
| # ------------------------------------------------------------- | ||
|
|
||
| function kron_rep( |
There was a problem hiding this comment.
I suggest this to be an internal function, can you rename this as _kron_rep
src/quantum_object.jl
Outdated
| return kron(xs...) | ||
| end | ||
|
|
||
| function kron_permutation(perm::Vector{Int}, dims::Vector{Int}) |
There was a problem hiding this comment.
I suggest this to be an internal function, can you rename this as _kron_permutation
test/quantum_objects.jl
Outdated
|
|
||
| @test permute(ket_abc, [2, 3, 1]) ≈ tensor(ket_b, ket_c, ket_a) | ||
| @test permute(bra_abc, [2, 3, 1]) ≈ tensor(bra_b, bra_c, bra_a) | ||
| @test permute(op_abc, [2, 3, 1]) ≈ tensor(op_b, op_c, op_a) |
There was a problem hiding this comment.
Can you also add the tests for raising potential errors ?
For example:
permute(ket_abc, [1, 2])
permute(ket_abc, [4, 5, 6])
test/quantum_objects.jl
Outdated
| @@ -291,4 +291,27 @@ | |||
| @test typeof(SparseMatrixCSC(Md).data) == SparseMatrixCSC{Int64,Int64} | |||
| @test typeof(SparseMatrixCSC(Ms).data) == SparseMatrixCSC{Int64,Int64} | |||
| @test typeof(SparseMatrixCSC{ComplexF64}(Ms).data) == SparseMatrixCSC{ComplexF64,Int64} | |||
|
|
|||
| # tensor tests | |||
There was a problem hiding this comment.
maybe make the comment
# tensor and permute tests
src/quantum_object.jl
Outdated
|
|
||
| function kron_rep( | ||
| dims; | ||
| labels="abcdefghijklmnopqrstuvwxyz"[1:length(dims)] |
There was a problem hiding this comment.
Is there a better way to define these labels for arbitrary length(dims)?
Because it seems that this doesn't work if the number of sub-systems is larger than 26 (a ~ z)
|
Could you also put the docstring of |
src/quantum_object.jl
Outdated
|
|
||
| # create string representations of the subsystems | ||
| # and store them in the order of the permutation | ||
| xs = [] |
There was a problem hiding this comment.
We should about to create empty lists of undefined types and then pushing to them. This is inefficient and allocates memory
src/quantum_object.jl
Outdated
| if length(order) != length(op.dims) | ||
| throw(ArgumentError("The order vector must have the same length as the number of subsystems")) | ||
| end | ||
| perm = kron_permutation(order, op.dims) |
There was a problem hiding this comment.
I don’t know how much inefficient is calling data[perm, perm], especially for sparse matrices
|
Hi @aarontrowbridge, thank you for this PR. I’m not sure if this is the most efficient and simple way to permute the dimensions. I would suggest to use reshaping and |
This hopefully resolves issue #95, it doesnt handle super operators, but perhaps that can be added in the future on top of this.