An Incremental Learning framework for Large-scale CTR Prediction

TL;DR

This paper presents an incremental learning framework for large-scale CTR prediction that accelerates training, adapts quickly to new data, and improves revenue and click-through rates in a real-world recommendation system.

Contribution

It introduces a novel incremental learning approach using a teacher-student paradigm for large-scale CTR prediction, enabling faster updates and better adaptation to new data.

Findings

12x faster training and deployment cycles

Consistent RPM lift across multiple traffic segments

Significant CTR increase on new items

Abstract

In this work we introduce an incremental learning framework for Click-Through-Rate (CTR) prediction and demonstrate its effectiveness for Taboola's massive-scale recommendation service. Our approach enables rapid capture of emerging trends through warm-starting from previously deployed models and fine tuning on "fresh" data only. Past knowledge is maintained via a teacher-student paradigm, where the teacher acts as a distillation technique, mitigating the catastrophic forgetting phenomenon. Our incremental learning framework enables significantly faster training and deployment cycles (x12 speedup). We demonstrate a consistent Revenue Per Mille (RPM) lift over multiple traffic segments and a significant CTR increase on newly introduced items.