Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[SYSTEMDS-2102] Vectorize gradients for Factorization Machines function #1803

Draft
wants to merge 2 commits into
base: main
Choose a base branch
from

Conversation

j143
Copy link
Contributor

@j143 j143 commented Apr 7, 2023

image

j143 and others added 2 commits April 7, 2023 21:42
CP°fcall°./nn/layers/fm.dml°backward°true°5°3°dout=_mVar779·MATRIX·FP64°X=X_batch·MATRIX·FP64°w0=w0·MATRIX·FP64°W=W·MATRIX·FP64°V=V·MATRIX·FP64°dw0°dW°dV
@j143
Copy link
Contributor Author

j143 commented Apr 7, 2023

the code doesn't seem to be correct:

-- old g_V2 --
0.000 0.000
0.000 -0.001
0.000 -0.000
0.000 -0.002
0.000 0.000
0.000 -0.001
0.000 -0.014

-- g_V2 --
0.000 -0.017

@Baunsgaard
Copy link
Contributor

LGTM

but remove the commented code.

@j143
Copy link
Contributor Author

j143 commented Apr 25, 2023

Hi @Baunsgaard - thanks for review.

I believe the vectorization I made is incorrect, especially output dimensions.
I will keep this in draft for a while.

@j143 j143 marked this pull request as draft April 25, 2023 14:09
@Baunsgaard
Copy link
Contributor

Hi @Baunsgaard - thanks for review.

I believe the vectorization I made is incorrect, especially output dimensions. I will keep this in draft for a while.

aww sad, it looked good at a surface glance, maybe we do not have enough tests ?
or is it that the NN is only in the Application tests again.

@j143
Copy link
Contributor Author

j143 commented Apr 29, 2023

We verify the with gradients at the moment. but, we need to verify against R implementation once.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
Status: Unplanned
2 participants