What is the explainability of machine learning? It is divided into two parts, one is the recommended material flow can be explained, the other is the model can be explained. Common stages of machine learning online link and the corresponding detailed introduction as well as the purpose of trace function 2, recommended materials can be explained, debug link: why to debug and common debug strategies 3, recalled materials can be seen, user behavior: It aims to reproduce all the behaviors of users at that time, including but not limited to macro behaviors such as clicking, browsing, adding purchase, and fine behaviors such as clicking the main image and browsing comments. Rich user features, item features, query features, end-to-side features and other features 5, real-time model quality analysis &&effect verification 6, shaply value on FLink ML: Interpretation for large models 1)Flink ML application model can be interpreted in a new way: shaply value model distributed way 2)Flink ML application model & Data integration new implementation way

Speakers:

Zhang Ying: Jingdong, Algorithm engineer, Zhang Ying, Alink, Flink AI Extend Contributor, architect of JD Intelligent Analysis Platform Department, mainly responsible for the real-time direction and algorithm interpretation system of JD Intelligent Analysis Platform Department, used to be responsible for the sample link of JD search and recommendation model, training and tuning of small model framework