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

Automatic scaling of sketching operators #70

Closed
rileyjmurray opened this issue Aug 22, 2023 · 2 comments
Closed

Automatic scaling of sketching operators #70

rileyjmurray opened this issue Aug 22, 2023 · 2 comments
Milestone

Comments

@rileyjmurray
Copy link
Contributor

Right now, our sparse operators have entries in {0,+1, -1}, and our dense operators are sampled from the standard normal distribution or the uniform distribution over [-1, 1]. The covariance matrices of these sketching operators will be scalar multiples of the identity matrix. However, the preferred scaling is that the covariance matrices are equal to the identity. We should update RandBLAS so that the correct scaling is automatically applied.

@rileyjmurray rileyjmurray changed the title Enhancement: automatic scaling of sketching operators Automatic scaling of sketching operators Jun 4, 2024
@rileyjmurray rileyjmurray added this to the 1.0 milestone Jun 5, 2024
@rileyjmurray
Copy link
Contributor Author

This will effectively be resolved by the isometry_scale_factor functions introduced in #104. (Although I still need to document those functions.)

@rileyjmurray
Copy link
Contributor Author

Resolved in #108, which adds an isometry_scale member to DenseDist and SparseDist. We don't apply this automatically though. The web docs speak to why in an FAQ.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant