altar.cuda.bayesian.cudaCoolingStep

Module Contents

Classes

class altar.cuda.bayesian.cudaCoolingStep.cudaCoolingStep(beta, theta, likelihoods, **kwds)

Encapsulation of the state of the calculation at some particular β value

beta

theta

prior

data

posterior

precision

classmethod start(cls, annealer)

Build the first cooling step by asking {model} to produce a sample set from its initializing prior, compute the likelihood of this sample given the data, and compute a (perhaps trivial) posterior

classmethod alloc(cls, samples, parameters, dtype)

Allocate storage for the parts of a cooling step

clone(self)

Make a new step with a duplicate of my state

computePosterior(self, batch=None)

(Re-)Compute the posterior from prior, data, and (updated) beta

copyFromCPU(self, step)

Copy cpu step to gpu step

copyToCPU(self, step)

copy gpu step to cpu step