Hyperbandit: Contextual Bandit with Hypernetwork for Time-Varying User Preferences in Streaming Recommendation

By Chenglei Shen et al.

Published on June 3, 2018

Read the original document by opening this link in a new tab.

Table of Contents

1. Abstract
2. Introduction
3. Problem Formulation
3.1 Bandit-based Streaming Recommendation
3.2 Time-Varying User Preferences
4. Hyperbandit: The Proposed Algorithm
4.1 Algorithm Overview
4.2 Hypernetwork Assisted Bandit Policy
5. Conclusion
References

Summary

The document discusses HyperBandit, a contextual bandit approach using hypernetwork, designed to address time-varying user preferences in streaming recommendation systems. It introduces the concept of time period embeddings and true user preference matrices to model dynamic user preferences. HyperBandit consists of a bandit policy for item recommendation and a hypernetwork for adapting user preferences based on time periods. The algorithm aims to capture periodic user preference changes efficiently and demonstrates superior performance in real-world streaming recommendation tasks.