Crate lfa

Expand description

lfa is a framework for online learning of linear function approximation models. Included is a suite of scalar and multi-output approximation helpers, a range of basis functions for feature construction, and a set of optimisation routines. The framework is designed to support fallible operations for which a custom error type is provided.

§Example

The code below illustrates how one might use lfa to fit a simple linear model with SGD.

extern crate lfa;
extern crate rand;
extern crate rand_distr;

use lfa::{
    LFA,
    basis::{Basis, Polynomial},
    optim::SGD,
};
use rand::{Rng, thread_rng};
use rand_distr::{Uniform, Normal};
use spaces::Space;

fn main() {
    const M: f64 = 1.0;
    const C: f64 = -0.5;

    let mut fa = LFA::scalar(
        Polynomial::new(1, 1).with_zeroth(),
        SGD(1.0)
    );

    let mut rng = thread_rng();
    let mut data_dist = Uniform::new_inclusive(-1.0, 1.0);
    let mut noise_dist = Normal::new(0.0, 0.01).unwrap();

    for x in rng.sample_iter(&data_dist).take(1000) {
        let y_exp = M*x + C;
        let y_sample = y_exp + rng.sample(noise_dist);

        fa.update_with(&[x], |_, y_pred| y_sample - y_pred).ok();
    }

    let rmse = (
        (fa.weights[0] - M).powi(2) +
        (fa.weights[1] - C).powi(2)
    ).sqrt();

    assert!(rmse < 1e-3);
}

Modules§

basis: Basis function representations used to generate Features.
optim: SGD-based weight optimisers.

Structs§

Error
LFA: Linear function approximator.
OutputIter: An iterator over the evaluated columns of a weight matrix.
SparseActivations: Internal type used to represent a sparse feature vector.

Enums§

ErrorKind
Features: Sparse/dense feature vector representation.

Type Aliases§