Recommendation systems are built to predict what users might like, especially when there are lots of choices available.
This post gives a deep dive into the architecture and issues experienced during the deployment of DLRM model. This algorithm was open-sourced by Facebook on 31st March 2019.
It’s a part of the popular MLPerf Benchmark.
Why should you consider using DLRM ?
This paper attempts to combine 2 important concepts that are driving the architectural changes in recommendation systems :