Lee and Seung's multiplicative update rule

ajensen1234 / nnmf-rs

Blazinging fast (!!!!!!) non-negative matrix factorization written in Rust!!!!!!!!

1 stars 0 forks source link

Open ajensen1234 opened 8 months ago

ajensen1234 commented 8 months ago

One aspect of this is that it runs element by element - might actually be a good use for some multithreading and parallelization. Maybe even GPU??

ajensen1234 commented 8 months ago

ajensen1234 commented 7 months ago

Implemented in #14 - linking against that branch and pulling into main.

ajensen1234 commented 7 months ago

Reopened for GPU

we can asynchronously define B,C,D,E in the convergence pipeline
we're going to need to write a BLAS (possibly) routine for matrix multiplication in order to keep things in the GPU for as long as possible