Struct linfa_svm::hyperparams::SvmParams

pub struct SvmParams<F: Float, T>(_);

Implementations

impl<F: Float, T> SvmParams<F, T>

pub fn new() -> Self

Create hyper parameter set

This creates a SvmParams and sets it to the default values:

• C values of (1, 1)
• Eps of 1e-7
• No shrinking
• Linear kernel

pub fn eps(self, new_eps: F) -> Self

Set stopping condition

This parameter controls the stopping condition. It checks whether the sum of gradients of the max violating pair is below this threshold and then stops the optimization proces.

pub fn shrinking(self, shrinking: bool) -> Self

Shrink active variable set

This parameter controls whether the active variable set is shrinked or not. This can speed up the optimization process, but may degredade the solution performance.

pub fn with_kernel_params(self, kernel: KernelParams<F>) -> Self

Set the kernel to use for training

This parameter specifies a mapping of input records to a new feature space by means of the distance function between any couple of points mapped to such new space. The SVM then applies a linear separation in the new feature space that may result in a non linear partitioning of the original input space, thus increasing the expressiveness of this model. To use the “base” SVM model it suffices to choose a Linear kernel.

pub fn with_platt_params(self, platt: PlattParams<F, ()>) -> Self

Set the platt params for probability calibration

pub fn gaussian_kernel(self, eps: F) -> Self

Sets the model to use the Gaussian kernel. For this kernel the distance between two points is computed as: d(x, x') = exp(-norm(x - x')/eps)

pub fn polynomial_kernel(self, constant: F, degree: F) -> Self

Sets the model to use the Polynomial kernel. For this kernel the distance between two points is computed as: d(x, x') = (<x, x'> + costant)^(degree)

pub fn linear_kernel(self) -> Self

Sets the model to use the Linear kernel. For this kernel the distance between two points is computed as : d(x, x') = <x, x'>

impl<F: Float, T> SvmParams<F, T>

pub fn pos_neg_weights(self, c_pos: F, c_neg: F) -> Self

Set the C value for positive and negative samples.

pub fn nu_weight(self, nu: F) -> Self

Set the Nu value for classification

The Nu value should lie in range [0, 1] and sets the relation between support vectors and solution performance.

impl<F: Float> SvmParams<F, F>

pub fn c_eps(self, c: F, eps: F) -> Self

Set the C value for regression

pub fn nu_eps(self, nu: F, eps: F) -> Self

Set the Nu-Eps value for regression

Trait Implementations

impl<F: Clone + Float, T: Clone> Clone for SvmParams<F, T>

fn clone(&self) -> SvmParams<F, T>

Returns a copy of the value.

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

Performs copy-assignment from source.

impl<F: Debug + Float, T: Debug> Debug for SvmParams<F, T>

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

Formats the value using the given formatter.

impl<F: Float, L> Default for SvmParams<F, L>

fn default() -> Self

Returns the “default value” for a type.

impl<F: Float, L> ParamGuard for SvmParams<F, L>

type Checked = SvmValidParams<F, L>

The checked hyperparameters

type Error = SvmError

Error type resulting from failed hyperparameter checking

fn check_ref(&self) -> Result<&Self::Checked, Self::Error>

Checks the hyperparameters and returns a reference to the checked hyperparameters if successful

fn check(self) -> Result<Self::Checked, Self::Error>

Checks the hyperparameters and returns the checked hyperparameters if successful

fn check_unwrap(self) -> Self::Checked

Calls check() and unwraps the result

impl<F: PartialEq + Float, T: PartialEq> PartialEq<SvmParams<F, T>> for SvmParams<F, T>

fn eq(&self, other: &SvmParams<F, T>) -> bool

This method tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &SvmParams<F, T>) -> bool

This method tests for !=.

impl<F: Float, T> StructuralPartialEq for SvmParams<F, T>