Project

General

Profile

Design #1460

Name of TensorMat; change to KroneckerProd?

Added by John Abbott almost 4 years ago. Updated over 3 years ago.

Status:
Closed
Priority:
Normal
Assignee:
John Abbott
Category:
Renaming
Target version:
CoCoALib-0.99800
Start date:
23 May 2020
Due date:
% Done:

100%

Estimated time:
1.33 h
Spent time:
1.35 h

Description

CoCoALib has a function called TensorMat.
The operation is also known as Kronecker product of matrices. The wikipedia entry suggests that KroneckerProd would be a better name.

What do you think?

History

#1 Updated by John Abbott almost 4 years ago

  • Subject changed from Name of TensorMat to Name of TensorMat; change to KroneckerProd?

I re-implemented the fn because I had forgotten (or didn't realise) that it was already in CoCoALib.
At the very least, there should be mention of "Kronecker product" in the documentation.

We could have both names. My current preference is to use the name KroneckerProd, and make TensorMat obsolescent.

#2 Updated by John Abbott almost 4 years ago

Another matter is where to put the defn of the function.
Currently TensorMat is defined in MatrixSpecial.C; I had expected to find it in MatrixOps.

Most of the other fns defined in MatrixSpecial are pseudo-ctors.
Is KroneckerProd a pseudo-ctor or an op? At the moment I think it is an op.

Maybe the original idea was to impl a pseudo-ctor where the original matrices would be stored (or pointed to), and then each entry computed on demand. The advantage is that almost no extra space is needed to store the KroneckerProd. This disadvantages are that the original matrices must be stored (at least via a ref count increment?) and that repeated access to an entry requires repeated computation of the value (just a multiplication). It seems hard to imagine a situation where the "lazy" impl really brings a real benefit (except perhaps for sparse matrices?)

At the moment I favour viewing it as an op (rather than a "smart" lazy ctor), so think that it is better moved to MatrixOps.

Comments?

#3 Updated by Anna Maria Bigatti almost 4 years ago

John Abbott wrote:

Another matter is where to put the defn of the function.
Currently TensorMat is defined in MatrixSpecial.C; I had expected to find it in MatrixOps.

I agree. I expect the function was older than the file.

Most of the other fns defined in MatrixSpecial are pseudo-ctors.
Is KroneckerProd a pseudo-ctor or an op? At the moment I think it is an op.

Maybe most of them are op's

#4 Updated by John Abbott over 3 years ago

  • Status changed from New to Feedback
  • Assignee set to John Abbott
  • % Done changed from 0 to 90

This has already been done.
TensorMat is now in obsolete.C

Was this already in 0.99710?

#5 Updated by John Abbott over 3 years ago

  • Status changed from Feedback to Closed
  • % Done changed from 90 to 100
  • Estimated time set to 1.33 h

This has been in feedback for a month or so. CLosing:

Conclusion: fn name changed KroneckerProd, impl creates new dense matrix, impl is in MatrixOps-KroneckerProd.C
Old fn TensorMat moved to obsolete.C

BTW KroneckerProd requires both mats to be over same ring (see issue #635).

PS I suppose I could have implemented the "view" version, and then made KroneckerProd use the view as initializer for a dense mat... (can't be bothered now)

Also available in: Atom PDF