scivex-optim

Optimization and numerical methods for Scivex. Root finding, minimization, integration, ODE solvers, linear programming, and curve fitting.

Highlights

• Root finding — Bisection, Newton-Raphson, Brent's method, secant method
• Minimization — Gradient descent, BFGS, L-BFGS-B, Nelder-Mead
• Linear programming — Revised simplex method for LP problems
• Curve fitting — Levenberg-Marquardt non-linear least squares
• Numerical integration — Trapezoidal, Simpson's, Gauss-Legendre quadrature
• ODE solvers — Euler, RK4, RK45, BDF2 for stiff systems
• PDE solvers — Wave equation (1D), Laplace equation (2D)
• Interpolation — 1D and 2D interpolation, B-splines
• Numerical differentiation — Forward, central, and Richardson extrapolation

Usage

use scivex_optim::prelude::*;

// Minimize Rosenbrock function
let f = |x: &Tensor<f64>| { /* Rosenbrock */ };
let grad = |x: &Tensor<f64>| { /* gradient */ };
let x0 = Tensor::from_vec(vec![-1.0, -1.0], &[2]);
let result = bfgs(f, grad, &x0, &MinimizeOptions::default()).unwrap();

// Root finding
let root = brent(|x| x * x - 2.0, 0.0, 2.0, 1e-10).unwrap();

// Numerical integration
let integral = simpson(|x| x.sin(), 0.0, std::f64::consts::PI, 100).unwrap();

License

MIT