`altar.models.seismic.cuda.cudaKinematicG`

Module Contents

Classes

class altar.models.seismic.cuda.cudaKinematicG.cudaKinematicG(name, locator, **kwds)

Bases: altar.cuda.models.cudaBayesian.cudaBayesian

KinematicG model with cuda

dataobs

default

doc = the observed data

green

doc = the name of the file with the Green functions

Nas

doc = number of patches along strike direction

Ndd

doc = number of patches along dip direction

Nmesh

doc = number of mesh points for each patch for fastsweeping

dsp

doc = the distance unit for each patch, in km

Nt

doc = number of time intervals for kinematic process

Npt

doc = number of mesh points for each time interval for fastsweeping

dt

doc = the time unit for each time interval (in s)

t0s

doc = the start time for each patch

cmodel

GF

gGF

gDprediction

cublas_handle

NGbparameters

gt0s

initialize(self, application): Initialize the state of the model given a {problem} specification

forwardModelBatched(self, theta, gf, prediction, batch, observation=None): KinematicG forward model in batch: cast Mb(x,y,t) :param theta: matrix (samples, parameters), sampling parameters :param gf: matrix (2*Ndd*Nas*Nt, observations), kinematicG green’s function :param prediction: matrix (samples, observations), the predicted data or residual between predicted and observed data :param batch: integer, the number of samples to be computed batch<=samples :param observation: matrix (samples, observations), duplicates of observed data :return: prediction as predicted data(observation=None) or residual (observation is provided)

forwardModel(self, theta, gf, prediction, observation=None): KinematicG forward model for single sample: cast Mb(x,y,t) :param theta: vector (parameters), sampling parameters :param gf: matrix (2*Ndd*Nas*Nt, observations), kinematicG green’s function :param prediction: vector (observations), the predicted data or residual between predicted and observed data :param observation: vector (observations), duplicates of observed data :return: prediction as predicted data(observation=None) or residual (observation is provided)

castSlipsOfTime(self, theta, Mb=None): Compute Mb (slips of patches over time) from a given set of parameters :param theta: a vector arranged in [slip (strike and dip), risetime, …] :param Mb: :return: Mb

linearGM(self, gf, Mb, prediction=None, observation=None): Perform prediction = Gb * Mb :param Gb: :param Mb: :param prediction: :return: prediction

cuEvalLikelihood(self, theta, likelihood, batch): Compute the likelihood from my forward problem

mergeCovarianceToGF(self): merge data covariance (cd) with green function

forwardProblem(self, application, theta=None): Perform the forward modeling with given {theta}