Crate numra 

Numra - Numerical Methods Library

Numra is a comprehensive numerical methods library for Rust.

This crate provides a unified interface to all Numra components:

• numra_core: Core traits and utilities (Scalar, Vector, Signal, Uncertainty)
• numra_linalg: Linear algebra (Matrix, LU decomposition)
• numra_nonlinear: Nonlinear solvers (Newton iteration)
• numra_ode: ODE solvers (DoPri5, Radau5, BDF, ESDIRK, Verner, etc.) and forward sensitivity analysis (ParametricOdeSystem, solve_forward_sensitivity)
• numra_sde: SDE solvers (Euler-Maruyama, Milstein, SRA1/SRA2)
• numra_dde: DDE solvers (Method of Steps)
• numra_fde: FDE solvers (L1 scheme for Caputo derivative)
• numra_ide: IDE solvers (Volterra, Prony series)
• numra_pde: PDE solvers (Method of Lines, moving boundaries)
• numra_spde: SPDE solvers (MOL + SDE steppers)

Sensitivity analysis — two distinct concepts

• compute_sensitivities / ParameterSensitivity / ParameterSensitivityResult (re-exported from numra-core): parameter-uncertainty / Sobol-style sensitivity for arbitrary scalar functions of uncertain inputs.
• solve_forward_sensitivity / solve_forward_sensitivity_with / ParametricOdeSystem / SensitivityResult (re-exported from numra-ode): forward sensitivity for ODE/DAE systems. Solves the augmented [y; S] system to compute S(t) = ∂y/∂p alongside the state trajectory.

Author: Moussa Leblouba Date: 9 February 2026 Modified: 6 May 2026

Re-exports

pub use numra_autodiff as autodiff;

pub use numra_dde as dde;

pub use numra_fde as fde;

pub use numra_fft as fft;

pub use numra_fit as fit;

pub use numra_ide as ide;

pub use numra_integrate as integrate;

pub use numra_interp as interp;

pub use numra_linalg as linalg;

pub use numra_nonlinear as nonlinear;

pub use numra_ocp as ocp;

pub use numra_ode as ode;

pub use numra_optim as optim;

pub use numra_pde as pde;

pub use numra_sde as sde;

pub use numra_signal as dsp;

pub use numra_spde as spde;

pub use numra_special as special;

pub use numra_stats as stats;

Modules

error

Error types for Numra.

scalar

Scalar type trait for numerical computation.

signal

Signal generators for forcing functions.

uncertainty

Uncertainty propagation and sensitivity analysis.

vector

Vector trait for numerical computation.

Structs

Interval

Interval arithmetic for bounds propagation.

ParameterSensitivity

Importance of one parameter for a scalar pure-function output.

ParameterSensitivityResult

Result of parameter-importance analysis on a scalar pure function.

SensitivityResult

Result of an ODE forward-sensitivity solve.

StateTransitionResult

Result of an initial-condition sensitivity solve.

Uncertain

An uncertain value with mean and variance.

Enums

LinalgError

Linear algebra specific errors.

NumraError

Errors that can occur in Numra operations.

Traits

ParametricOdeSystem

An ODE system parameterised by a parameter vector p.

Scalar

A real scalar type suitable for numerical computation.

Signal

A time-dependent signal for forcing functions.

Vector

A vector type for numerical computation.

Functions

compute_sensitivities

Compute finite difference sensitivities.

solve_forward_sensitivity

Compute the forward sensitivity matrix S(t) = ∂y(t) / ∂p of an ODE solution using any Solver.

solve_forward_sensitivity_with

Closure-shaped convenience wrapper around solve_forward_sensitivity.

solve_initial_condition_sensitivity

Compute the state-transition matrix Φ(t) = ∂y(t) / ∂y₀ of the linearised flow of dy/dt = f(t, y), using any Solver.

solve_initial_condition_sensitivity_with

Closure-shaped convenience wrapper around solve_initial_condition_sensitivity.

Type Aliases

NumraResult

Result type alias for Numra operations.