Hacker News new | ask | show | jobs
by orasis 871 days ago
I built a contextual bandit combining XGBoost with Thompson Sampling you can check out at https://improve.ai
1 comments

What's the added value over Thomson's sampling?
It can learn faster and generalize learning to unseen variants by learning the impact different features have on the conversion rate.

It can also learn how different variants perform in different contexts.