Multi Arm Bandit - Githubissues

jmaberk / RGPUCB

This code runs Bayesian optimization with the randomised Gaussian process upper confidence bound acquisition function

MIT License

5 stars 2 forks source link

Multi Arm Bandit #1

Open garitrik opened 3 years ago

garitrik commented 3 years ago

Can this package be used to solve multi-arm bandit problems?

jmaberk commented 3 years ago

Hi garitrik

This is a package for Bayesian optimisation, which can be viewed as a sort of continuous multi-armed bandit problem. However, it relies on correlation between neighbouring points which varies by distance. Hence it is not suited for standard multi-arm bandit problems. You are likely better off finding an algorithm designed for them.

garitrik commented 3 years ago

OK thank you.