fastmath.kernel.vector — generateme/fastmath 3.0.0-alpha1

Various vector based kernels

anova clj

(anova)

(anova {:keys [sigma k d] :or {sigma 1.0 k 1.0 d 1.0}})

Anova kernel.

Parameters:

• :sigma - multiplier (default: 1.0)
• :k and :d - exponents (default: 1.0)

b-spline clj

(b-spline)

(b-spline {:keys [n] :or {n 2.0}})

B-spline kernel with degree :n (default: 2.0).

bessel clj

(bessel)

(bessel {:keys [sigma n v distance]
         :or {sigma 1.0 n 2.0 v -1.0 distance v/dist}})

Bessel (of the first kind) kernel

Parameters:

• :sigma - shape (default: 1.0)
• :n - exponent factor (default: 2.0)
• :v - Bessel J order - 1 (default: -1.0)
• :distance - distance function (default: euclidean)

bessel2 clj

(bessel2)

(bessel2 {:keys [sigma order degree distance]
          :or {sigma 1.0 order 0.0 degree 1.0 distance v/dist}})

Bessel (of the first kind) kernel, R kernlab implementation.

Parameters:

• :sigma - shape (default: 1.0)
• :degree - exponent (default: 1.0)
• :order - Bessel J order (default: 0.0)
• :distance - distance function (default: euclidean)

cauchy clj

(cauchy)

(cauchy {:keys [sigma distance] :or {sigma 1.0 distance v/dist}})

Cauchy kernel.

Parameters:

• sigma - scale (default: 1.0)
• :distance - distance function (default: euclidean)

chi-square clj

(chi-square)

(chi-square _)

Chi-square kernel.

chi-square2 clj

(chi-square2)

(chi-square2 _)

Chi-square kernel, second version

dirichlet clj

(dirichlet)

(dirichlet {:keys [n] :or {n 1.0}})

Dirichlet kernel with :n dimensionality (default: 1.0).

exponential clj

(exponential)

(exponential {:keys [sigma distance] :or {sigma 1.0 distance v/dist}})

Exponential kernel.

Parameters:

• :sigma - shape of the kernel (default: 1.0)
• :distance - distance function (default: euclidean)

gaussian clj

(gaussian)

(gaussian {:keys [sigma distance] :or {sigma 1.0 distance v/dist}})

Gaussian kernel.

Parameters:

• :sigma - shape of the kernel (default: 1.0)
• :distance - distance function (default: euclidean)

generalized-histogram clj

(generalized-histogram)

(generalized-histogram {:keys [p] :or {p 2.0}})

Generalized histogram with :p exponent (default: 2.0).

generalized-t-student clj

(generalized-t-student)

(generalized-t-student {:keys [p distance]})

Generalized t-student.

Parameters:

• :p - exponent
• :distance - distance function (default: euclidean)

geometric clj

(geometric)

(geometric {:keys [n r distance] :or {r 1.0 distance v/dist}})

Geometric Compactly Supported kernel

Parameters:

• :n - dimension
• :r - shape (default: 1.0)
• :distance - distance function (default: euclidean)

Specific kernel names for :n:

• 1 - triangular
• 2 - circular
• 3 - spherical

hellinger clj

(hellinger)

(hellinger _)

Hellinger kernel.

histogram clj

(histogram)

(histogram _)

Histogram kernel.

hyperbolic-secant clj

(hyperbolic-secant)

(hyperbolic-secant {:keys [a distance] :or {a 1.0 distance v/dist}})

Hyperbolic secant kernel.

Parameters:

• :a scaling factor (default: 1.0)
• :distance - distance function (default: euclidean)

hyperbolic-tangent clj

(hyperbolic-tangent)

(hyperbolic-tangent {:keys [alpha c] :or {alpha 1.0 c 0.0}})

Hyperbolic tangent of the dot product.

Parameters:

• :alpha - dot product multiplier (default: 1.0)
• :c - shift (default: 0.0)

inverse-multiquadratic clj

(inverse-multiquadratic)

(inverse-multiquadratic {:keys [c distance] :or {c 1.0 distance v/dist}})

Inverse multiquadratic kernel.

Parameters:

:c - shift (default: 1.0) :distance - distance function (default: euclidean)

laplacian clj

(laplacian)

(laplacian {:keys [sigma distance] :or {sigma 1.0 distance v/dist}})

Laplacian kernel.

Parameters:

• :sigma - shape of the kernel (default: 1.0)
• :distance - distance function (default: euclidean)

linear clj

(linear)

(linear _)

Dot product kernel

log clj

(log)

(log {:keys [p distance]})

Logarithmic.

Parameters:

• :p - exponent (default: 2.0)
• :distance - distance function (default: euclidean)

matern clj

(matern)

(matern {:keys [order theta distance] :or {order 1 theta 1.0 distance v/dist}})

Matern kernel.

Parameters:

• :order - order of the kernel, should be odd (default: 1).
• :theta - shape (default: 1.0)
• :distance - distance function (default: euclidean)

Order of the Bessel K function is a half of :order parameter. For example to get Matern 5/2 kernel, call (matern 5).

multiquadratic clj

(multiquadratic)

(multiquadratic {:keys [c distance] :or {c 1.0 distance v/dist}})

Multiquadratic kernel.

Parameters:

:c - shift (default: 1.0) :distance - distance function (default: euclidean)

pearson clj

(pearson)

(pearson {:keys [sigma omega distance]
          :or {sigma 1.0 omega 1.0 distance v/dist}})

Pearson VII kernel

Parameters:

• :sigma - scale (default: 1.0)
• :omega - exponent (default: 1.0)
• :distance - distance function (default: euclidean)

periodic clj

(periodic)

(periodic {:keys [sigma periodicity distance]
           :or {sigma 1.0 periodicity 1.0 distance v/dist}})

Periodic kernel.

Parameters:

• :sigma - scale (default: 1.0)
• :periodicity - periodicity (default: 1.0)
• :distance - distance function (default: euclidean)

polynomial clj

(polynomial)

(polynomial {:keys [alpha c p] :or {alpha 1.0 c 0.0 p 2.0}})

Polynomial (power of dot product).

Parameters:

• :alpha - dot product multiplier (default: 1.0)
• :c - shift (default: 0.0)
• :p - exponent (default: 2.0)

power clj

(power)

(power {:keys [p distance] :or {p 2.0 distance v/dist}})

Power (negative) of distance.

Parameters:

• :p - exponent (default: 2.0)
• :distance - distance function (default: euclidean)

rational-quadratic clj

(rational-quadratic)

(rational-quadratic {:keys [c distance] :or {c 1.0 distance v/dist}})

Rational quadratic kernel.

Parameters:

:c - shift (default: 1.0) :distance - distance function (default: euclidean)

rbf->kernel clj

(rbf->kernel rbf-kernel)

(rbf->kernel rbf-kernel distance)

Convert RBF kernel as vector kernel using a distance function (default: euclidean).

spline clj

(spline)

(spline _)

Spline kernel.

wave clj

(wave)

(wave {:keys [sigma distance] :or {sigma 1.0 distance v/dist}})

Wave (sinc) kernel.

Parameters:

• :sigma - scale (default: 1.0)
• :distance - distance function (default: euclidean)