ThotDjehuty
d780ed81d7
Adds 9 new top-level / sub-modules to the Rust API only (no Python
bindings yet), with at least one analytic unit test per module.
New Rust modules:
- optimal_control::matrix_riccati (RK4 backward solver)
- timeseries_utils::nonsync_covariance (Hayashi-Yoshida)
- timeseries_utils::wavelet (Haar / Daubechies DWT and MODWT)
- risk_measures (VaR, CVaR, projected sub-gradient CVaR minimisation)
- graph::laplacian + graph::spectral_clustering (Jacobi + k-means++)
- topology (Vietoris-Rips persistent homology, bottleneck distance)
- volterra (Caputo Adams, Markovian lift, second-kind Volterra,
Fourier inversion of characteristic functions)
- signatures (truncated tensor signature, log-sig, random reservoir,
Salvi-Cass-Lyons signature kernel, shuffle product)
All previously stable APIs untouched; abi3-py38 ABI preserved.
New module tests: 29/29 passing. Pre-existing 5 unrelated failures
unchanged.
85 lines
2.4 KiB
Rust
85 lines
2.4 KiB
Rust
//! Optimal Control Module
|
||
//! ======================
|
||
//!
|
||
//! Generic optimal control algorithms for dynamical systems.
|
||
//!
|
||
//! # Features
|
||
//!
|
||
//! - Generic HJB PDE solver with finite differences
|
||
//! - Viscosity solutions for non-smooth value functions
|
||
//! - Upwind schemes for numerical stability
|
||
//! - Parallel processing support
|
||
//!
|
||
//! # Mathematical Foundation
|
||
//!
|
||
//! ## Hamilton-Jacobi-Bellman Equation
|
||
//!
|
||
//! For a general stochastic process dX_t = μ(X_t)dt + σ(X_t)dW_t,
|
||
//! the HJB equation is:
|
||
//!
|
||
//! ρV(x) = sup_u [μ(x,u)V'(x) + (σ²(x,u)/2)V''(x) + L(x,u)]
|
||
//!
|
||
//! where:
|
||
//! - V(x) is the value function
|
||
//! - ρ is the discount rate
|
||
//! - u is the control
|
||
//! - L(x,u) is the running cost
|
||
//!
|
||
//! ## Viscosity Solutions
|
||
//!
|
||
//! Handle non-smooth value functions via:
|
||
//! 1. Finite difference discretization
|
||
//! 2. Upwind schemes for stability
|
||
//! 3. Iterative convergence to viscosity solution
|
||
|
||
pub mod backtest;
|
||
pub mod hjb_solver;
|
||
pub mod jump_diffusion;
|
||
pub mod kalman_filter;
|
||
#[cfg(feature = "python-bindings")]
|
||
pub mod kalman_py_bindings;
|
||
pub mod matrix_riccati;
|
||
pub mod mrsjd;
|
||
pub mod ou_estimator;
|
||
#[cfg(feature = "python-bindings")]
|
||
pub mod py_bindings;
|
||
pub mod regime_switching;
|
||
pub mod viscosity;
|
||
|
||
pub use hjb_solver::{HJBConfig, HJBResult, HJBSolver};
|
||
pub use matrix_riccati::{solve_matrix_riccati, RiccatiConfig, RiccatiResult};
|
||
pub use jump_diffusion::{
|
||
JumpDiffusionConfig, JumpDiffusionResult, JumpDiffusionSolver, JumpDistribution,
|
||
};
|
||
pub use kalman_filter::{
|
||
FilterResult, KalmanFilter, KalmanState, LinearObservation, LinearStateTransition,
|
||
ObservationModel, RTSSmoother, SmootherResult, StateTransitionModel, UnscentedKalmanFilter,
|
||
};
|
||
pub use mrsjd::{MRSJDConfig, MRSJDResult, MRSJDSolver, RegimeJumpParameters};
|
||
pub use regime_switching::{
|
||
RegimeParameters, RegimeSwitchingConfig, RegimeSwitchingResult, RegimeSwitchingSolver,
|
||
};
|
||
pub use viscosity::{ViscosityConfig, ViscositySolver};
|
||
|
||
use thiserror::Error;
|
||
|
||
#[derive(Error, Debug)]
|
||
pub enum OptimalControlError {
|
||
#[error("Numerical error: {0}")]
|
||
NumericalError(String),
|
||
|
||
#[error("Convergence failed: {0}")]
|
||
ConvergenceError(String),
|
||
|
||
#[error("Invalid parameters: {0}")]
|
||
InvalidParameters(String),
|
||
|
||
#[error("Insufficient data: need at least {0} points")]
|
||
InsufficientData(usize),
|
||
|
||
#[error("Matrix computation error: {0}")]
|
||
MatrixError(String),
|
||
}
|
||
|
||
pub type Result<T> = std::result::Result<T, OptimalControlError>;
|