Sampling

This notebook illustrates the different probability distributions that can be used to generate datasets of physical parameters.

import os
import sys

sys.path.insert(0, os.path.join(".."))

import numpy as np
import matplotlib.pyplot as plt

from infobs.sampling import samplers
from infobs.sampling import Mixture

Predefined samplers

samplers.__all__  # Sampler is the abstract class from which all samplers inherit

['Sampler', 'Constant', 'Uniform', 'LogUniform', 'BoundedPowerLaw']

cst = samplers.Constant(1.0)
cst.get(5)

array([1., 1., 1., 1., 1.])

unf = samplers.Uniform(0.0, 1.0)
unf.get(5)

array([0.39079479, 0.026211  , 0.56833878, 0.19770359, 0.52410246])

logunf = samplers.LogUniform(1, 10)
logunf.get(5)

array([7.30769182, 4.33391842, 1.17338091, 4.69312246, 8.30467105])

bdpl = samplers.BoundedPowerLaw(0.5, 1, 10)
bdpl.get(5)

array([6.79175107, 3.53320021, 3.82415888, 4.71669006, 4.05413621])

n_samples = 100_000
bins = 50

plt.figure(dpi=150)

plt.subplot(2, 2, 1)
x = cst.get(n_samples)
plt.hist(x, bins)

plt.subplot(2, 2, 2)
x = unf.get(n_samples)
plt.hist(x, bins)

plt.subplot(2, 2, 3)
x = logunf.get(n_samples)
plt.hist(x, bins)

plt.subplot(2, 2, 4)
x = bdpl.get(n_samples)
plt.hist(x, bins)

plt.tight_layout()
plt.show()

New samplers from existing ones

bdpl2 = bdpl.copy_other_bounds(10, 100)
bdpl2.get(5)

array([26.80674634, 53.89577313, 13.49510575, 31.95299392, 40.06931273])

plt.figure(dpi=150, figsize=(0.5 * 6.4, 0.5 * 4.8))

x = bdpl2.get(n_samples)
plt.hist(x, bins)

plt.show()

Defining your own samplers

class Exponential(samplers.Sampler):
    def __init__(self, lamda: float):
        """
        TODO
        """
        assert lamda > 0

        self.lamda = lamda

    def get(self, n: int) -> np.ndarray:
        """
        TODO
        """
        return np.random.exponential(self.lamda, n)

exp = Exponential(1.0)
exp.get(5)

array([0.59646487, 1.16991262, 0.79159345, 0.16810737, 0.72334426])

plt.figure(dpi=150, figsize=(0.5 * 6.4, 0.5 * 4.8))

x = exp.get(n_samples)
plt.hist(x, bins)

plt.show()

Samplers mixture

mxt = Mixture([unf, bdpl], [1, 5])

mxt.get(5)

array([4.05883558, 7.8928761 , 1.00402175, 9.12008436, 1.31337906])

plt.figure(dpi=150, figsize=(0.5 * 6.4, 0.5 * 4.8))

x = mxt.get(n_samples)
plt.hist(x, bins)

plt.show()