Nonlinear data transformer using Gaussian kernel.

gaussian_transformer.py

import numpy as np

class GaussianTransformer(object):
    def __init__(self, sigma=1.0, b=100, random_state=1):
        self.sigma = sigma
        self.b = 100
        self.random_state = random_state

    def fit(self, X):
        b = np.min([X.shape[0], self.b])
        r = np.random.RandomState(self.random_state)
        idx = r.permutation(X.shape[0])
        self.Xce = X[idx[:b]]
        self.Xce2 = np.c_[np.sum(self.Xce**2, axis=1)]
        return self

    def fit_transform(self, X):
        self.fit(X)
        return self.transform(X)

    def transform(self, X):
        X2 = np.c_[np.sum(X**2, axis=1)]
        dist2 = X2 + self.Xce2.T - 2*X.dot(self.Xce.T)
        return np.exp(-dist2/(2*self.sigma**2))