Quickstart

1. Call from rusampling import Ru.
2. Create an instance of the Ru class, passing the log-pdf logf, the dimension d (default 1), and any extra arguments that the log-pdf takes.
3. Call this object's rvs method, passing the number of samples n to generate.

So, if f(x) is a pdf in 1 dimension, take 100 samples, by passing the log-pdf logf(x):

t = Ru(log_f)
samples = t.rvs(n=100)

An example: sampling from the gamma distribution, passing args that gamma.logpdf take to Ru:

from scipy.stats import gamma

t = Ru(gamma.logpdf, a=2)
samples = t.rvs(n=50)

An example: parameter sampling from the normal distribution with prior \(1/\sigma\), using some data:

from scipy.stats import norm
import numpy as np

def log_posterior(params, x):
    mu = params[0]
    sigma = params[1]
    log_f = np.sum(norm.logpdf(x, loc=mu, scale=sigma)) 
    log_prior = -np.log(sigma)
    return log_f + log_prior

x_normal = [0.95368014, -0.65628003, -0.63509242,  
            1.34746566,  2.0896044, 0.97076896]

t = Ru(log_posterior, x=x_normal, d=2)
parameter_samples = t.rvs(n=200)