COI — Common Optimization Interfaces¶
Geoff (Generic Optimization Framework & Frontend) is the project of bringing numerical optimization, machine learning and reinforcement learning to the operation of particle accelerators. It consists of several packages, usually under the prefix “cernml”.
This package defines Common Optimization Interfaces that facilitate using numerical optimization and reinforcement learning (RL) on the same optimization problems. This makes it possible to unify both approaches into a generic optimization application.
To get started, please read Quickstart.
The cernml-coi-utils package provides many additional features that complement the COIs.
This repository can be found online on CERN’s Gitlab.
- Tutorials
- User Guide
- Quickstart
- The Core API
- Making Your Code Findable
- Custom Per-Problem Optimizers
- Making an Optimization Configurable via GUI
- Control Flow of Optimization Problems
- Waiting for New Data Without Blocking the GUI
- More Optimization Interfaces
- Optimizing Points on an LSA Function
- Migration Guide for COI v0.9
- Code Examples
- API Reference
- Roadmap
- Changelog