Small Companies Can Now Access Deep-Learning Recommendation Models with Low-Memory ROBE Array

ROBE Array allows small businesses to access AI

A breakthrough low-memory technique by Rice University computer scientists could put one of the most resource-intensive forms of artificial intelligence–deep-learning recommendation models (DLRM)–within reach of small companies.

DLRM systems are a form of AI which learns how to make relevant suggestions for users. Top-of-the line training models require more than 100 terabytes in memory and supercomputer processing. They are only available to technology giants who have deep pockets.

Rice’s \”random off block embedding Array,\” or ROBE array, could change this. This algorithmic approach is designed to reduce the size of DLRM structures known as embedding table memory structures. It will be presented at the Conference on Machine Learning and Systems 2022 in Santa Clara (California) this week, where it won Outstanding Paper awards.

Source:
https://techxplore.com/news/2022-08-robe-array-small-companies-access.html