API Reference#

PyOED consists of several modular subpackages which can be employed with great flexibility. An important aspect of PyOED’s design is that every class is associated with a validated, introspectable configurations dataclass — see Key Concepts for details.

Find What You Need — by Workflow#

Not sure where to start in the API? Use the table below to jump directly to the relevant subpackage for the task you want to perform.

Task	Primary subpackage	Supporting subpackage
Define a forward (simulation) model	Models	Configs
Define observation noise / prior distribution	Models — Error Models	Stats
Define an observation operator (sensor map)	Models — Observation Operators	Models
Solve an inverse problem (3D-Var, Kalman, etc.)	Data Assimilation	Models
Run sensor-placement OED (A/D-optimal)	OED — Sensor Placement	OED Criteria
Run robust OED (uncertain hyperparameters)	OED — Robust OED	OED
Optimise a design (SciPy, binary, REINFORCE)	Optimization	OED
Time-dependent inversion (4D-Var, smoothing)	Data Assimilation — Smoothing	Models
Sample from a distribution / run MCMC	Statistics	Models — Error Models
Extend PyOED (add a new model or criterion)	Architecture	Configs
Add a custom error model or prior	Models — Error Models	Configs
Add a custom observation operator	Models — Observation Operators	Models Core
Add a custom sampler or proposal	Statistics — Sampling	Statistics Core

Glossary#

A definition of all key terms (OED, A-optimality, EIG, 3D-Var, sensor placement, REINFORCE, …) is available in the Glossary.