Struct linfa_kernel::KernelParams

pub struct KernelParams<F> { /* fields omitted */ }

Defines the set of prameters needed to build a kernel

Implementations

impl<F: Float> KernelParams<F>

pub fn method(mut self: Self, method: KernelMethod<F>) -> KernelParams<F>

Setter for method. Can be chained with kind and transform.

Arguments

• method: The inner product that will be used by the kernel

Returns

The modified set of params

Example

use linfa_kernel::{Kernel, KernelMethod};
use linfa::traits::Transformer;
use ndarray::Array2;

let data = Array2::from_shape_vec((3,2), vec![1., 2., 3., 4., 5., 6.,]).unwrap();

// Build a kernel from `data` with the defaul parameters
// and then set the preferred method
let params = Kernel::params().method(KernelMethod::Linear);
let kernel = params.transform(&data);

pub fn kind(mut self: Self, kind: KernelType) -> KernelParams<F>

Setter for kind. Can be chained with method and transform.

Arguments

• kind: The kind of kernel to build, either dense or sparse

Returns

The modified set of params

Example

use linfa_kernel::{Kernel, KernelType};
use linfa::traits::Transformer;
use ndarray::Array2;

let data = Array2::from_shape_vec((3,2), vec![1., 2., 3., 4., 5., 6.,]).unwrap();

// Build a kernel from `data` with the defaul parameters
// and then set the preferred kind
let params = Kernel::params().kind(KernelType::Dense);
let kernel = params.transform(&data);

Trait Implementations

impl<'a, F: Float> Transformer<&'a ArrayBase<OwnedRepr<F>, Dim<[usize; 2]>>, Kernel<ArrayBase<ViewRepr<&'a F>, Dim<[usize; 2]>>>> for KernelParams<F>

pub fn transform(&self, x: &'a Array2<F>) -> Kernel<ArrayView2<'a, F>>

Builds a kernel from the input data without copying it.

A reference to the input data will be kept by the kernel through an ArrayView

Parameters

• x: matrix of records (##records, ##features) in input

Returns

A kernel build from x according to the parameters on which this method is called

Panics

If the kernel type is Sparse and the number of neighbors specified is not between 1 and ##records-1

impl<'a, F: Float, T: Targets> Transformer<&'a DatasetBase<ArrayBase<OwnedRepr<F>, Dim<[usize; 2]>>, T>, DatasetBase<Kernel<ArrayBase<ViewRepr<&'a F>, Dim<[usize; 2]>>>, &'a T>> for KernelParams<F>

pub fn transform(
    &self,
    x: &'a DatasetBase<Array2<F>, T>
) -> DatasetBase<Kernel<ArrayView2<'a, F>>, &'a T>

Builds a new Dataset with the kernel as the records and the same targets as the input one.

A reference to the input records will be kept by the kernel through an ArrayView

Parameters

• x: A dataset with a matrix of records (##records, ##features) and any targets

Returns

A new dataset with:

• records: a kernel build from x.records() according to the parameters on which this method is called
• targets: same as x.targets()

Panics

If the kernel type is Sparse and the number of neighbors specified is not between 1 and ##records-1

impl<'a, F: Float, T: Targets> Transformer<&'a DatasetBase<ArrayBase<ViewRepr<&'a F>, Dim<[usize; 2]>>, T>, DatasetBase<Kernel<ArrayBase<ViewRepr<&'a F>, Dim<[usize; 2]>>>, &'a [<T as Targets>::Elem]>> for KernelParams<F>

pub fn transform(
    &self,
    x: &'a DatasetBase<ArrayView2<'a, F>, T>
) -> DatasetBase<Kernel<ArrayView2<'a, F>>, &'a [T::Elem]>

Builds a new Dataset with the kernel as the records and the same targets as the input one.

A reference to the input records will be kept by the kernel through an ArrayView

Parameters

• x: A dataset with a matrix of records (##records, ##features) and any targets

Returns

A new dataset with:

• records: a kernel build from x.records() according to the parameters on which this method is called
• targets: a slice of x.targets()

Panics

If the kernel type is Sparse and the number of neighbors specified is not between 1 and ##records-1

impl<'a, F: Float> Transformer<ArrayBase<ViewRepr<&'a F>, Dim<[usize; 2]>>, Kernel<ArrayBase<ViewRepr<&'a F>, Dim<[usize; 2]>>>> for KernelParams<F>

pub fn transform(&self, x: ArrayView2<'a, F>) -> Kernel<ArrayView2<'a, F>>

Builds a kernel from a view of the input data.

A reference to the input data will be kept by the kernel through an ArrayView

Parameters

• x: view of a matrix of records (##records, ##features)

A kernel build from x according to the parameters on which this method is called

Panics

If the kernel type is Sparse and the number of neighbors specified is not between 1 and ##records-1