Uber has developed Two-Tower Embeddings (TTE) for its recommendation systems, using embeddings to improve scalability and efficiency. TTE provides personalized retrieval from a large pool of stores and can be used for final ranking in recommendation systems. The TTE model utilizes engagement data and localized relations to optimize training and inference. It offers feature extensibility and has been successful in improving performance and lowering costs at Uber. Challenges include model size, training time, and evaluation metrics.
Table of contents
IntroductionWhat are Embeddings and Two-Tower Embeddings (TTE)?A Closer Look: Problem and MotivationDeep Dive: The Embeddings SolutionModelingChallengesResult and TakeawaysWhat’s Next?Acknowledgements12 Impressions