Yoel Zeldes • 8/21/2018

Uncertainty for CTR Prediction: One Model to Clarify Them All

This technical article details Taboola's implementation of a neural network that simultaneously predicts an item's Click-Through Rate (CTR) and quantifies the uncertainty of that prediction. It covers the model's architecture, including modules for item, context, fusion, and estimation, and explains how it handles data, model, and measurement uncertainty in a principled way for improved recommendations.

0 comments

#Machine Learning #Neural Networks #Recommender Systems