Struct TrainingSettings

pub struct TrainingSettings {
    pub accuracy_error: f64,
    pub step: f64,
    pub max_iterations: u64,
    pub error_predictions: usize,
    pub reset_interval: u64,
    pub elasticity: f64,
    pub debug: bool,
}

Stores training settings.

Fields

accuracy_error: f64

Acceptable accuracy in error.

step: f64

The minimum step value.

When error_predictions is set to 0, this is used as fixed step.

max_iterations: u64

Maximum number of iterations.

error_predictions: usize

The number of error prediction terms.

More terms accelerate the search, but might lead to instability.

reset_interval: u64

The interval to reset error predictions, in case they are far off or become unstable.

elasticity: f64

A factor greater than zero to prevent under or over-stepping.

E.g. 0.95

This is used because predicted errors does not provide information about the gradient directly in the domain. Elasticity is used to estimate the gradient.

debug: bool

Whether to print out result each reset interval.

Trait Implementations

impl Clone for TrainingSettings

fn clone(&self) -> TrainingSettings

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for TrainingSettings

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

Formats the value using the given formatter.

impl Copy for TrainingSettings