Add kron(): kronecker product for 2D matrices #1039
Conversation
| /// | ||
| /// The kronecker product of a LxN matrix A and a MxR matrix B is a (L*M)x(N*R) | ||
| /// matrix K formed by the block multiplication A_ij * B. | ||
| pub fn kron<T>(a: &Array2<T>, b: &Array2<T>) -> Array2<T> |
There was a problem hiding this comment.
To include this we'd need to use the same kind of argument type we usually use for reading an array - an &ArrayBase like other places that take any kind of array.
Minor style issues - I'd suggest it can have the slightly longer name kronecker. Also note that ndarray uses the element type parameter A by convention, so we need to keep that to keep it consistent. Thanks
There was a problem hiding this comment.
I think kron would align with numpy, as a datapoint, just wanted to mention. It also is an operation usually done on 2 dimensional matrices, though I believe numpy supports higher dimensional kronecker products.
| .into_iter() | ||
| .zip(a.iter()) | ||
| { | ||
| let v: Array2<T> = Array2::from_elem((dimbr, dimbc), *(elem)) * b; |
There was a problem hiding this comment.
Here we'd like to avoid creating an intermediate array. We'd just use Zip instead of from_elem and assign. Other improvements are less important (alloc uninit array instead of zeroed, other ways to loop?). We would however like to replace the other zip with Zip because the iteration order of exact_chunks_mut is not well specified.
|
@notmgsk Are you interested in continuing to work on this? If not I will probably build off of this PR and respond to the review. |
|
I went ahead and made #1105, which I think responds to the feedback here. |
|
Superseded by #1105 |
Closes #652
H/T @ethanhs