Trait Optimizable 

pub trait Optimizable {
    type Inputs;
    type Targets;

    // Required method
    fn compute_grad(
        &self,
        params: &[f64],
        inputs: &Self::Inputs,
        targets: &Self::Targets,
    ) -> (f64, Vec<f64>);
}

Trait for models which can be gradient-optimized.

Required Associated Types

type Inputs

The input data type to the model.

type Targets

The target data type to the model.

Required Methods

fn compute_grad( &self, params: &[f64], inputs: &Self::Inputs, targets: &Self::Targets, ) -> (f64, Vec<f64>)

Compute the gradient for the model.

Dyn Compatibility

This trait is dyn compatible.

In older versions of Rust, dyn compatibility was called "object safety".

Implementors

impl Optimizable for BaseLogisticRegressor

Computing the gradient of the underlying Logistic Regression model.

The gradient is given by

X^T(h(Xb) - y) / m

where h is the sigmoid function and b the underlying model parameters.

type Inputs = Matrix<f64>

type Targets = Vector<f64>

impl Optimizable for LinRegressor

type Inputs = Matrix<f64>

type Targets = Vector<f64>

impl<'a, T: Criterion> Optimizable for BaseNeuralNet<'a, T>

Compute the gradient of the Neural Network using the back propagation algorithm.

type Inputs = Matrix<f64>

type Targets = Matrix<f64>