BoTorch Tutorials

The tutorials here will help you understand and use BoTorch inyour own work. They assume that you are familiar with bothBayesian optimization (BO) and PyTorch.

• If you are new to BO, we recommend you start with theAx docs and thefollowingtutorial paper.
• If you are new to PyTorch, the easiest way to get started iswith theWhat is PyTorch?tutorial.

Using BoTorch with Ax

For practitioners who are interested in running experimentsto optimize various objectives using Bayesian optimization,we recommend usingAx rather than BoTorch.Ax provides a user-friendly interface forexperiment configuration and orchestration, while choosing anappropriate Bayesian optimization algorithm to optimize thegiven objective, following BoTorch best practices.

For researchers who are interested in running experiments withtheir custom BoTorch models and acquisition functions,Ax's Modular BoTorch Interface offers a convenientway to leverage custom BoTorch objects while utilizingAx experiment configuration and orchestration. Check outModular BoTorch tutorialto learn how to use custom BoTorch objects in Ax!Seethis documentationfor additional information.

Full Optimization Loops

In some situations (e.g. when working in a non-standard setting,or if you want to understand and control various details of theBO loop), then you may also consider working purely in BoTorch.The tutorials in this section illustrate this approach.

Bite-Sized Tutorials

Rather than guiding you through full end-to-end BO loops, thetutorials in this section focus on specific tasks that you willencounter in customizing your BO algorithms. For instance, youmay want towrite a custom acquisition functionand thenuse a custom zero-th order optimizerto optimize it.

Advanced Usage

Tutorials in this section showcase more advanced ways of usingBoTorch. For instance,this tutorialshows how to perform BO if your objective function is an image,by optimizing in the latent space of a variational auto-encoder(VAE).