Metropolis-Hastings Sampler¶

class sampyl.Metropolis(logp, start, tune_interval=100, **kwargs)¶

Metropolis-Hastings sampler for drawing from a distribution defined by a logp function.

Has automatic scaling such that acceptance rate stays around 50%

Parameters:

logp – function log P(X) function for sampling distribution.
start – Dictionary of starting state for the sampler. Should have one element for each argument of logp.
scale – scalar or 1D array-like. initial scaling factor for proposal distribution.
tune_interval – int.
scale – scalar or 1D array-like. initial scaling factor for proposal distribution.
tune_interval – int. number of samples between tunings of scale factor.

Example:

def logp(x, y):
    ...

start = {'x': x_start, 'y': y_start}
metro = sampyl.Metropolis(logp, start)
chain = metro.sample(20000, burn=5000, thin=4)

sample(num, burn=0, thin=1, n_chains=1, progress_bar=True)¶

Sample from \(P(X)\)

Parameters:

num – int. Number of samples to draw from \(P(X)\).
burn – (optional) int. Number of samples to discard from the beginning of the chain.
thin – (optional) float. Thin the samples by this factor.
n_chains – (optional) int. Number of chains to return. Each chain is given its own process and the OS decides how to distribute the processes.
progress_bar – (optional) boolean. Show the progress bar, default = True.

Returns:

Record array with fields taken from arguments of logp function.

Sampyl