Practical Multi-Task Learning for Rare Conversions in Ad Tech
By: Yuval Dishi , Ophir Friedler , Yonatan Karni and more
Potential Business Impact:
Helps ads show up for the right people.
We present a Multi-Task Learning (MTL) approach for improving predictions for rare (e.g., <1%) conversion events in online advertising. The conversions are classified into "rare" or "frequent" types based on historical statistics. The model learns shared representations across all signals while specializing through separate task towers for each type. The approach was tested and fully deployed to production, demonstrating consistent improvements in both offline (0.69% AUC lift) and online KPI performance metric (2% Cost per Action reduction).
Similar Papers
No One Left Behind: How to Exploit the Incomplete and Skewed Multi-Label Data for Conversion Rate Prediction
Machine Learning (CS)
Helps online ads show the right products to more people.
Personalized Product Search Ranking: A Multi-Task Learning Approach with Tabular and Non-Tabular Data
Information Retrieval
Finds you better stuff when you shop online.
Efficient Multi-Task Learning via Generalist Recommender
Information Retrieval
Makes online suggestions work better for more people.