search

nickgreenquist / cu2rec

CUDA Implementation of Parallel Matrix Factorization Algorithm for Recommender Systems
GNU Affero General Public License v3.0
13 stars 1 forks source link

Put user bias into shared memory in SGD #23

Closed dorukkilitcioglu closed 5 years ago

dorukkilitcioglu commented 5 years ago

User bias is small and can fit in shared memory. Should make things faster.

dorukkilitcioglu commented 5 years ago

Decided against this as the gain wasn't high enough, because the user bias was only accessed once per kernel call.