Machine Learning Intermediate Pro

Contextual Bandits for Personalization

LinUCB, neural bandits, and action features — the workhorse algorithm behind notification and content personalization

30 min read 19 views

Learn how contextual bandits extend multi-armed bandits by conditioning the policy on features of the user and situation: the LinUCB algorithm and why a linear reward model gives closed-form confidence bounds, neural/deep contextual bandits for richer reward functions, action features vs. arm identity for scaling to large or changing catalogues, and position bias in bandit-served lists.

Part of the Reinforcement Learning & Long-term Optimization path →

Preview Subscribe to unlock

Practice questions (5)

Justify a Contextual Bandit for Send-Time Personalization

Intermediate · Free

View →
Compute LinUCB Scores by Hand for Two Arms

Intermediate

View →
Design Cold Start for a Large, Turning-Over Creative Catalogue

Intermediate

View →
Diagnose and Fix Position Bias in a Bandit-Ranked Carousel

Advanced

View →
Evaluate a Proposed Offline Test of a New Bandit Policy

Advanced

View →

Quizzes (1)

Contextual Bandits for Personalization Quiz

0 attempts · 10 min · Pass 70% · Intermediate

Take quiz →