[−][src]Crate liblbfgs

Limited memory BFGS (L-BFGS) algorithm ported from liblbfgs

Example

```// 0. Import the lib
use liblbfgs::lbfgs;

const N: usize = 100;

// 1. Initialize data
let mut x = [0.0 as f64; N];
for i in (0..N).step_by(2) {
x[i] = -1.2;
x[i + 1] = 1.0;
}

// 2. Defining how to evaluate function and gradient
let evaluate = |x: &[f64], gx: &mut [f64]| {
let n = x.len();

let mut fx = 0.0;
for i in (0..n).step_by(2) {
let t1 = 1.0 - x[i];
let t2 = 10.0 * (x[i + 1] - x[i] * x[i]);
gx[i + 1] = 20.0 * t2;
gx[i] = -2.0 * (x[i] * gx[i + 1] + t1);
fx += t1 * t1 + t2 * t2;
}

Ok(fx)
};

let prb = lbfgs()
.with_max_iterations(5)
.with_orthantwise(1.0, 0, 99) // enable OWL-QN
.minimize(
&mut x,                   // input variables
evaluate,                 // define how to evaluate function
|prgr| {                  // define progress monitor
println!("iter: {:}", prgr.niter);
false                 // returning true will cancel optimization
}
)
.expect("lbfgs owlqn minimize");

println!("fx = {:}", prb.fx);```

Modules

 line Find a satisfactory step length along predefined search direction math Backend for lbfgs vector operations

Structs

 LbfgsParam L-BFGS optimization parameters. LbfgsState LBFGS optimization state allowing iterative propagation Orthantwise Orthant-Wise Limited-memory Quasi-Newton (OWL-QN) algorithm Problem Represents an optimization problem. Progress Store optimization progress data, for progress monitor Report

Functions

 default_evaluate Default test function (rosenbrock) adopted from liblbfgs sample.c default_progress Default progress monitor adopted from liblbfgs sample.c lbfgs