Struct LinearRegression 

pub struct LinearRegression<T: RealNumber> { /* private fields */ }

Represents a linear regression model.

The LinearRegression struct implements a linear regression model for predicting a target variable based on one or more input features. It uses the least squares method to estimate the weights of the linear model.

Type Parameters

• T: The numeric type used for calculations. Must implement the RealNumber trait.

Fields

• weights: The weights of the logistic regression model, with the first being the bias weight.

Examples

use rusty_ai::regression::linear::LinearRegression;
use rusty_ai::data::dataset::Dataset;
use nalgebra::{DMatrix, DVector};

// Create a new linear regression model
let mut model = LinearRegression::<f64>::new();

// Fit the model to a dataset
let x = DMatrix::from_row_slice(3, 2, &[1.0, 2.0, 3.0, 4.0, 5.0, 6.0]);
let y = DVector::from_vec(vec![1.5, 2.5, 3.5]);
let dataset = Dataset::new(x, y);
let learning_rate = 0.01;
let max_steps = 1000;
let epsilon = Some(0.001);
let progress = Some(100);
let result = model.fit(&dataset, learning_rate, max_steps, epsilon, progress);

// Make predictions using the trained model
let x_test = DMatrix::from_row_slice(2, 2, &[1.0, 2.0, 3.0, 4.0]);
let predictions = model.predict(&x_test);
assert!(predictions.is_ok());

Implementations

impl<T: RealNumber> LinearRegression<T>

pub fn new() -> Self

Creates a new LinearRegression model with default weights.

The default weights are initialized to 1.0 for each feature, including the bias weight.

pub fn with_params( dimension: Option<usize>, weights: Option<DVector<T>>, ) -> Result<Self, Box<dyn Error>>

Creates a new LinearRegression model with custom parameters.

Arguments

• dimension: The dimension of the input features. If None, the dimension will be inferred from the provided weights.
• weights: The initial weights for the linear regression model. If None, default weights will be used.

Returns

A Result containing the LinearRegression model if the parameters are valid, or an error message if the parameters are invalid.

Errors

An error will be returned if:

• Both dimension and weights are None.
• The length of weights is not equal to dimension + 1 to account for the bias weight.

pub fn weights(&self) -> &DVector<T>

A reference to the weights of the linear regression model.

pub fn predict(&self, x_pred: &DMatrix<T>) -> Result<DVector<T>, Box<dyn Error>>

Makes predictions using the trained linear regression model.

Arguments

• x_pred: The input features for which to make predictions.

Returns

A Result containing the predicted target values if successful, or an error message if an error occurs during prediction.

pub fn fit( &mut self, dataset: &Dataset<T, T>, lr: T, max_steps: usize, epsilon: Option<T>, progress: Option<usize>, ) -> Result<String, Box<dyn Error>>

Fits the linear regression model to a dataset.

Arguments

• dataset: The dataset containing the input features and target values.
• lr: The learning rate for gradient descent.
• max_steps: The maximum number of steps to perform during training.
• epsilon: The convergence threshold. If the change in weights is below this threshold, training will stop.
• progress: The number of steps at which to display progress information. If None, no progress information will be displayed.

Returns

A Result containing a success message if training is successful, or an error message if an error occurs during training.

Errors

An error will be returned if:

• The number of steps for progress visualization is 0.
• The gradient turns to NaN during training.

Trait Implementations

impl<T: Clone + RealNumber> Clone for LinearRegression<T>

fn clone(&self) -> LinearRegression<T>

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl<T: Debug + RealNumber> Debug for LinearRegression<T>

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl<T: RealNumber> Default for LinearRegression<T>

fn default() -> Self

Creates a new LinearRegression model with default weights.

The default weights are initialized to 1.0 for each feature, including the bias weight.

impl<T: RealNumber> RegressionMetrics<T> for LinearRegression<T>

fn mse( &self, y_true: &DVector<T>, y_pred: &DVector<T>, ) -> Result<T, Box<dyn Error>>

Computes the mean squared error (MSE) between the true values and the predicted values.

fn mae( &self, y_true: &DVector<T>, y_pred: &DVector<T>, ) -> Result<T, Box<dyn Error>>

Computes the mean absolute error (MAE) between the true values and the predicted values.

fn r2( &self, y_true: &DVector<T>, y_pred: &DVector<T>, ) -> Result<T, Box<dyn Error>>

Computes the coefficient of determination (R^2) between the true values and the predicted values.