Summary of Lirank: Industrial Large Scale Ranking Models at Linkedin, by Fedor Borisyuk et al.
LiRank: Industrial Large Scale Ranking Models at LinkedIn
by Fedor Borisyuk, Mingzhou Zhou, Qingquan Song, Siyu Zhu, Birjodh Tiwana, Ganesh Parameswaran, Siddharth Dangi, Lars Hertel, Qiang Xiao, Xiaochen Hou, Yunbo Ouyang, Aman Gupta, Sheallika Singh, Dan Liu, Hailing Cheng, Lei Le, Jonathan Hung, Sathiya Keerthi, Ruoyan Wang, Fengyu Zhang, Mohit Kothari, Chen Zhu, Daqi Sun, Yun Dai, Xun Luan, Sirou Zhu, Zhiwei Wang, Neil Daftary, Qianqi Shen, Chengming Jiang, Haichao Wei, Maneesh Varshney, Amol Ghoting, Souvik Ghosh
First submitted to arxiv on: 10 Feb 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary A novel large-scale ranking framework, LiRank, is introduced by LinkedIn, combining state-of-the-art modeling architectures and optimization methods to produce high-performance models. This framework includes improvements such as Residual DCN, which adds attention and residual connections to the DCNv2 architecture. The authors also share insights on combining and tuning these architectures to create a unified model, using techniques like Dense Gating, Transformers, and Residual DCN. Additionally, novel calibration methods are proposed, along with strategies for training and compressing models through quantization and vocabulary compression. The deployment setup for large-scale use cases of Feed ranking, Jobs Recommendations, and Ads click-through rate (CTR) prediction is detailed. A/B tests are used to summarize the learnings, highlighting the most effective technical approaches, which have led to significant improvements in metrics across the board at LinkedIn. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary LiRank is a new way for LinkedIn to rank things like job listings or ads. It combines lots of different ideas from other researchers and makes them work together really well. They also came up with some new ways to make sure the rankings are good, and they found ways to make the models smaller so they can be used on more devices. This is important because it means people will see better results when they search for jobs or look at ads. The researchers tested these ideas and found that they work really well, making users happy! |
Keywords
* Artificial intelligence * Attention * Optimization * Quantization
