Skip to content

RBFSampler

Extracts random features which approximate an RBF kernel.

This is a powerful way to give non-linear capacity to linear classifiers. This method is also called "random Fourier features" in the literature.

Parameters

  • gamma

    Default1.0

    RBF kernel parameter in (-gamma * x^2).

  • n_components

    Default100

    Number of samples per original feature. Equals the dimensionality of the computed feature space.

  • seed

    Typeint | None

    DefaultNone

    Random number seed.

Examples

from river import feature_extraction as fx
from river import linear_model as lm
from river import optim
from river import stream

X = [[0, 0], [1, 1], [1, 0], [0, 1]]
Y = [0, 0, 1, 1]

model = lm.LogisticRegression(optimizer=optim.SGD(.1))

for x, y in stream.iter_array(X, Y):
    model.learn_one(x, y)
    y_pred = model.predict_one(x)
    print(y, int(y_pred))
0 0
0 0
1 0
1 1

model = (
    fx.RBFSampler(seed=3) |
    lm.LogisticRegression(optimizer=optim.SGD(.1))
)

for x, y in stream.iter_array(X, Y):
    model.learn_one(x, y)
    y_pred = model.predict_one(x)
    print(y, int(y_pred))
0 0
0 0
1 1
1 1

Methods

learn_one

Update with a set of features x.

A lot of transformers don't actually have to do anything during the learn_one step because they are stateless. For this reason the default behavior of this function is to do nothing. Transformers that however do something during the learn_one can override this method.

Parameters

  • x'dict'

transform_one

Transform a set of features x.

Parameters

  • x'dict'
  • y — defaults to None

Returns

dict: The transformed values.