• Typically
  • Batch
    • Pre-computed
    • Decouple compute from serving
    • Lower operational load
    • Make prediction for everyone
  • Real-time
    • Responsive to time-sensitive context
    • Reduced cost on non-visiting users
    • Make prediction per request
    • Higher cost on operation
• In this talk, we will focus on real-time aka on-demand

Overview

Taobao

Facebook

Taobao - ATBRG

Doordash

Retrieval

The Youtube Recommendations (DNN)

Instagram (word2vec)

Ranker

The Youtube Recommendations (via DNN)

Alibaba (via Transformer)

Offline-to-online transitions

Related

• Billion-scale Comodity Embedding for E-commerce Recommendation an AlibabaBillion-scale Comodity Embedding for E-commerce Recommendation an Alibaba
  
  
  
  
  
  
  
  
  Reference
  
  J. Wang, P. Huang, H. Zhao, Z. Zhang, B. Zhao, and D. L. Lee, “Billion-scale Commodity Embedding for E-commerce Recommendation in Alibaba,” arXiv:1803.02349 [cs], May 2018, Acc...
• Embedding-based Retrieval in Facebook SearchEmbedding-based Retrieval in Facebook Search
  
  
  
  
  Reference
  
  J.-T. Huang et al., “Embedding-based Retrieval in Facebook Search,” Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pp. 2553–2...
• ATBRG - Adaptive Target-Behavior Relational Graph Network for Effective RecommendationATBRG - Adaptive Target-Behavior Relational Graph Network for Effective Recommendation
  Note
  
  
  
  
  
  Reference
  
  Y. Feng, B. Hu, F. Lv, Q. Liu, Z. Zhang, and W. Ou, “ATBRG: Adaptive Target-Behavior Relational Graph Network for Effective Recommendation,” Proceedings of the 43rd Internati...
• Deep Neural Networks for YouTube RecommendationsDeep Neural Networks for YouTube Recommendations
  
  
  Retrieval
  
  
  
  Ranker
  
  
  
  
  
  Reference
  
  P. Covington, J. Adams, and E. Sargin, “Deep Neural Networks for YouTube Recommendations,” in Proceedings of the 10th ACM Conference on Recommender Systems, ...
• Behavior Sequence Transformer for E-commerce Recommendation in AlibabaBehavior Sequence Transformer for E-commerce Recommendation in Alibaba
  Note
  
  
  
  
  
  Reference
  
  Q. Chen, H. Zhao, W. Li, P. Huang, and W. Ou, “Behavior Sequence Transformer for E-commerce Recommendation in Alibaba,” arXiv:1905.06874 [cs], May 2019, Accessed: Nov. 06, 20...
• DoorDash: Understanding Search Intent with Better Recall
• Powered by AI - Instagrams explore recommender system

Reference:

• Eugene Yan - System Design for Recommendations and Search - MLOps Community
• Keynote 7: Moving Beyond Recommender Models - Even Oldridge (NVIDIA), Karl Byleen-Higley (NVIDIA)

#recommender-system #recommender-system/archtecture #search