Crate picard 

Picard

Fast Independent Component Analysis using preconditioned L-BFGS optimization.

This crate implements the Picard algorithm from:

Pierre Ablin, Jean-François Cardoso, Alexandre Gramfort. “Faster independent component analysis by preconditioning with Hessian approximations” IEEE Transactions on Signal Processing, 2018

Example

use picard::{Picard, PicardConfig};
use ndarray::Array2;

// Generate some test data (n_features x n_samples)
let x = Array2::<f64>::zeros((10, 1000));

// Fit ICA with default settings
let result = Picard::fit(&x)?;

// Or with custom configuration
let config = PicardConfig::builder()
    .n_components(5)
    .max_iter(200)
    .ortho(true)
    .build();
let result = Picard::fit_with_config(&x, &config)?;

// Access results
let sources = &result.sources;
let unmixing = &result.unmixing;

Re-exports

pub use ndarray;

Modules

utils

Utility functions for ICA analysis.

Structs

ConfigBuilder

Builder for constructing PicardConfig with a fluent API.

Cube

Cubic density.

Exp

Exponential density.

Picard

The PICARD Independent Component Analysis solver.

PicardConfig

Configuration parameters for the PICARD algorithm.

PicardResult

Result of running the PICARD algorithm.

Tanh

Hyperbolic tangent density.

Enums

DensityType

Enumeration of built-in density types.

PicardError

Errors that can occur during PICARD computation.

Traits

Density

Trait for density functions used in ICA.