hpr_analysis/lib.rs
1//! Monte Carlo dispersion, sensitivity analysis, optimization and challenge specifications.
2//!
3//! [Monte Carlo dispersion][guide-mc] shows a run and how far to trust it. [Sensitivity
4//! analysis][guide-sa] ranks which inputs matter. [Optimization][guide-opt] finds the design that
5//! hits a target. [Start here][guide-start] says what works today and what is planned.
6//!
7//! [guide-mc]: https://hpr.fusionspace.co/monte-carlo.html
8//! [guide-sa]: https://hpr.fusionspace.co/sensitivity.html
9//! [guide-opt]: https://hpr.fusionspace.co/optimization.html
10//! [guide-start]: https://hpr.fusionspace.co/start-here.html
11//! [roadmap]: https://github.com/nrdptel/fusionspace-eridanus/blob/main/docs/ROADMAP.md
12//!
13//! - [`montecarlo`]: one rocket flown many times with its uncertain inputs drawn afresh each
14//! time. It's seeded and reproducible sample by sample, and runs in parallel with the
15//! `parallel` feature.
16//! - [`statistics`]: the spread of what the samples gave, with the failed ones counted.
17//! - [`ellipse`]: where the landings scatter, as an ellipse holding a chosen share of them.
18//! - [`sensitivity`]: which uncertain inputs move a result most, by Morris's screening and
19//! Sobol' indices.
20//! - [`optimize`]: the design variables that make a result as small as it can be, to hit a target
21//! apogee, say. It uses CMA-ES, NSGA-II for two or more goals at once, or EGO for a slow model.
22//!
23//! Status: dispersion, the apogee's spread, landing ellipses and sensitivity analysis work. So
24//! does optimization over continuous and discrete variables, with constraints and with one goal
25//! or several. Optimizing a Monte Carlo statistic (a robust mode) and challenge specifications are
26//! planned ([the roadmap][roadmap]).
27
28pub mod ellipse;
29mod error;
30pub mod montecarlo;
31pub mod optimize;
32pub mod sensitivity;
33pub mod statistics;
34pub mod table;
35
36pub use error::AnalysisError;