Source code for skpro.density

import abc

import numpy as np
from scipy.integrate import simps
from sklearn.base import BaseEstimator
from sklearn.neighbors import KernelDensity

from .utils import not_existing


[docs]def ecdf(a): """ Returns the empirical distribution function of a sample Parameters ---------- a: array Input array representing a sample Returns ------- array xs Empirical cdf of the input sample array ys """ xs = np.sort(np.array(a)) ys = np.arange(1, len(xs) + 1) / float(len(xs)) return xs, ys
[docs]def step_function(xs, ys): """ Returns a step function from x-y pair sample Parameters ---------- xs x values ys corresponding y values Returns ------- function A step function """ def func(x): index = np.searchsorted(xs, x) index = len(ys) - 1 if index >= len(ys) else index return ys[index] return func
[docs]class DensityAdapter(BaseEstimator, metaclass=abc.ABCMeta): """ Abtract base class for density adapter that transform an input into an density cdf/pdf interface """ @abc.abstractmethod def __call__(self, inlet): """ Adapter entry point Parameters ---------- mixed inlet Input for the adapter transformation """ raise NotImplementedError()
[docs] @abc.abstractmethod def pdf(self, x): """ Probability density function Parameters ---------- x Returns ------- mixed Density function evaluated at x """ raise NotImplementedError()
[docs] @abc.abstractmethod def cdf(self, x): """ Cumulative density function Parameters ---------- x Returns ------- mixed Cumulative density function evaluated at x """ raise NotImplementedError()
[docs]class KernelDensityAdapter(DensityAdapter): """ DensityAdapter that uses kernel density estimation to transform samples """ def __init__(self, estimator=KernelDensity()): self.estimator = estimator def __call__(self, sample): """ Adapter entry point Parameters ---------- np.array(M) inlet: Sample of length M """ # fit kernel density estimator self._min_sample = min(sample) self._max_sample = max(sample) self._std_sample = np.std(sample) self.estimator.fit(sample[:, np.newaxis])
[docs] def cdf(self, x): a = 10 grid_size = 1000 minus_inf = self._min_sample - a * self._std_sample step = (x - minus_inf) / grid_size grid = np.arange(minus_inf, x, step) pdf_estimation = np.exp(self.estimator.score_samples(grid.reshape(-1, 1))) integral = simps(y=pdf_estimation, dx=step) return integral
[docs] def pdf(self, x): x = np.array(x) try: return np.exp(self.estimator.score_samples(x))[0] except ValueError: return np.exp(self.estimator.score_samples(x.reshape(-1, 1)))[0]
[docs]class EmpiricalDensityAdapter(DensityAdapter): """ DensityAdapter that uses empirical cdf to transform samples """ def __init__(self): self.xs_ = None self.ys_ = None self.step_function_ = None def __call__(self, sample): """ Adapter entry point Parameters ---------- np.array(M) inlet: Bayesian sample of length M """ self.xs_, self.ys_ = ecdf(sample) self.step_function_ = step_function(self.xs_, self.ys_)
[docs] def cdf(self, x): return self.step_function_(x)
[docs] @not_existing def pdf(self, x): pass