search

boostorg / ublas

Boost.uBlas
https://www.boost.org/doc/libs/release/libs/numeric/ublas
111 stars 152 forks source link

Matrix/Tensor Factorization #66

Closed thomasyang1207 closed 5 years ago

thomasyang1207 commented 5 years ago

Operations such as matrix multiplication are BLAS level 3 operations, while matrix addition is level 1. Suppose we have the following computation of 4 matrices:

A C + B D;

Suppose A = B, but the user of the library may or may not know this. Would it be a good idea to find a way to compute this instead:

A (C+ D)

I think this is a rather trivial topic, but I thought it might be cool for uBLAS to have this

yimyom commented 5 years ago

there was a project on optimizing the template tree with interesting results. It was done by an excellent student in 2014.

It would be nice to revive this project and try to extend it. Are you interested? The project is here: https://github.com/BoostGSoC14/boost.ublas

On Sat, Mar 30, 2019 at 2:21 PM Thomas notifications@github.com wrote:

Operations such as matrix multiplication are BLAS level 3 operations, while matrix addition is level 1. Suppose we have the following computation of 4 matrices:

A C + B D;

Suppose A = B, but the user of the library may or may not know this. Would it be a good idea to find a way to compute this instead:

A (C+ D)

I think this is a rather trivial topic, but I thought it might be cool for uBLAS to have this

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/boostorg/ublas/issues/66, or mute the thread https://github.com/notifications/unsubscribe-auth/ACDlcPrXj2gQ2aEuXROpx5DJ2zZFXi06ks5vbtgmgaJpZM4cTgZ4 .

bassoy commented 5 years ago

Hi all,

we are now trying to do this with Boost/Yap which is very convenient to use and has already a builtin transform function.

Best Cem

David Bellot notifications@github.com schrieb am Mo., 1. Apr. 2019, 11:21:

there was a project on optimizing the template tree with interesting results. It was done by an excellent student in 2014.

It would be nice to revive this project and try to extend it. Are you interested? The project is here: https://github.com/BoostGSoC14/boost.ublas

On Sat, Mar 30, 2019 at 2:21 PM Thomas notifications@github.com wrote:

Operations such as matrix multiplication are BLAS level 3 operations, while matrix addition is level 1. Suppose we have the following computation of 4 matrices:

A C + B D;

Suppose A = B, but the user of the library may or may not know this. Would it be a good idea to find a way to compute this instead:

A (C+ D)

I think this is a rather trivial topic, but I thought it might be cool for uBLAS to have this

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/boostorg/ublas/issues/66, or mute the thread < https://github.com/notifications/unsubscribe-auth/ACDlcPrXj2gQ2aEuXROpx5DJ2zZFXi06ks5vbtgmgaJpZM4cTgZ4

.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/boostorg/ublas/issues/66#issuecomment-478503077, or mute the thread https://github.com/notifications/unsubscribe-auth/ADzxpcl4Ct5bds1K5fbHlwigssMCnu-Iks5vcc-mgaJpZM4cTgZ4 .