abcpmc Package

sampler Module

class abcpmc.sampler.Sampler(N, Y, postfn, dist, threads=1, pool=None)

ABC population monte carlo sampler

Parameters:

• N – number of particles
• Y – observed data set
• postfn – model function (a callable), which creates a new dataset x for a given theta
• dist – distance function rho(X, Y) (a callable)
• threads – (optional) number of threads. If >1 and no pool is given <threads> multiprocesses will be started
• pool – (optional) a pool instance which has a <map> function

close()

Tries to close the pool (avoid hanging threads)

particle_proposal_cls

alias of ParticleProposal

sample(prior, eps_proposal)

Launches the sampling process. Yields the intermediate results per iteration.

Parameters:

• prior – instance of a prior definition (or an other callable) see sampler.GaussianPrior
• eps_proposal – an instance of a threshold proposal (or an other callable) see sampler.ConstEps

Yields pool:

yields a namedtuple representing the values of one iteration

class abcpmc.sampler.GaussianPrior(mu, sigma)

Normal gaussian prior

Parameters:

• mu – scalar or vector of means
• sigma – scalar variance or covariance matrix

class abcpmc.sampler.TophatPrior(min, max)

Tophat prior

Parameters:

• min – scalar or array of min values
• max – scalar or array of max values

class abcpmc.sampler.ParticleProposal(sampler, eps, pool, kwargs)

Creates new particles using twice the weighted covariance matrix (Beaumont et al. 2009)

class abcpmc.sampler.OLCMParticleProposal(sampler, eps, pool, kwargs)

Bases: abcpmc.sampler.ParticleProposal

Creates new particles using an optimal loacl covariance matrix (Fillipi et al. 2012)

class abcpmc.sampler.KNNParticleProposal(sampler, eps, pool, kwargs)

Bases: abcpmc.sampler.ParticleProposal

Creates new particles using a covariance matrix from the K-nearest neighbours (Fillipi et al. 2012) Set k as key-word arguement in abcpmc.Sampler.particle_proposal_kwargs

abcpmc.sampler.weighted_cov(values, weights)

Computes a weighted covariance matrix

Parameters:

• values – the array of values
• weights – array of weights for each entry of the values

Returns sigma:

the weighted covariance matrix

abcpmc.sampler.weighted_avg_and_std(values, weights, axis=None)

Return the weighted avg and standard deviation.

Parameters:

• values – Array with the values
• weights – Array with the same shape as values containing the weights
• axis – (optional) the axis to be used for the computation

Returns avg, sigma:  

weighted average and standard deviation

threshold Module

Various different threshold implementations

Created on Jan 19, 2015

author: jakeret

class abcpmc.threshold.ConstEps(T, eps)

Bases: abcpmc.threshold.EpsProposal

Constant threshold. Can be used to apply alpha-percentile threshold decrease :param eps: epsilon value

class abcpmc.threshold.EpsProposal(T)

Bases: object

next()

reset()

class abcpmc.threshold.ExponentialConstEps(max, min, T1, T2)

Bases: abcpmc.threshold.EpsProposal

class abcpmc.threshold.ExponentialEps(T, max, min)

Bases: abcpmc.threshold.EpsProposal

Exponentially decreasing threshold

Parameters:

• max – epsilon at t=0
• min – epsilon at t=T
• T – number of iterations

class abcpmc.threshold.LinearConstEps(max, min, T1, T2)

Bases: abcpmc.threshold.EpsProposal

Linearly decreasing threshold until T1, then constant until T2

Parameters:

• max – epsilon at t=0
• min – epsilon at t=T
• T1 – number of iterations for decrease
• T2 – number of iterations for constant behavior

class abcpmc.threshold.LinearEps(T, max, min)

Bases: abcpmc.threshold.EpsProposal

Linearly decreasing threshold

Parameters:

• max – epsilon at t=0
• min – epsilon at t=T
• T – number of iterations

class abcpmc.threshold.ListEps(T, eps_vals)

Bases: abcpmc.threshold.EpsProposal

mpi_util Module

A helper util for message passing interface (MPI) handlings

class abcpmc.mpi_util.MpiPool(mapFunction=<built-in function map>)

Bases: object

isMaster()

Returns true if the rank is 0

map(function, sequence)

abcpmc.mpi_util.mpiBCast(value)

Mpi bcasts the value and returns the value from the master (rank = 0).