Method

Event sequence encoder

Self-supervised embeddings of user event sequences (CoLES and similar).

17 cases from 15 companies apply Event sequence encoder. 4 were checked in an online A/B test; 3 report results in production.

Typical tasks: Learn user and item embeddings, Recommend from what the user just did, Predict churn and lifetime value, Decide who gets an offer and 10 more. Companies: Skoltech, VTB, Ramp, Affirm, Airbnb, Avito and 9 more. Often combined with: Self-supervised learning, Gradient boosting, Masked language modeling, Contrastive learning, Multi-task learning and 31 more.