linna.main

Module Contents

Functions

ml_sampler(outdir, theory, priors, data, cov, init, pool, nwalkers, gpunode, omegab2cut=None, nepoch=4500, method='zeus', nbest=None, chisqcut=None, loglikelihoodfunc=None)

LINNA main function with hyperparameters set to values described in To et al. 2022

ml_sampler_core(ntrainArr, nvalArr, nkeepArr, ntimesArr, ntautolArr, meanshiftArr, stdshiftArr, outdir, theory, priors, data, cov, init, pool, nwalkers, device, dolog10index, ypositive, temperatureArr, omegab2cut=None, docuda=False, tsize=1, gpunode=None, nnmodel_in=None, params=None, method='emcee', nbest=None, chisqcut=None, loglikelihoodfunc=None, nsigma=3)

LINNA main function
linna.main.ml_sampler(outdir, theory, priors, data, cov, init, pool, nwalkers, gpunode, omegab2cut=None, nepoch=4500, method='zeus', nbest=None, chisqcut=None, loglikelihoodfunc=None)[source]

LINNA main function with hyperparameters set to values described in To et al. 2022

Parameters

  • outdir (string) – output directory

  • theory (function) – theory model

  • str (priors (dict of) – [float, float]): string can be either flat or gauss. If the string is ‘flat’, [a,b] indicates the lower and upper limits of the prior. If the string is ‘gauss’, [a,b] indicates the mean and sigma.

  • data (1d array) – float array, data vector

  • cov (2d array) – float array, covariance matrix

  • init (ndarray) – initial guess of mcmc,

  • pool (mpi pool, optional) – a mpi pool instance that can do pool.map(function, iterables).

  • nwalkers (int) –

  • gpunode (string) – name of gpu node

  • omegab2cut (list of int) – 2 elements containing the lower and upper limits of omegab*h^2

  • nepoch (int, optional) – maximum number of epoch for the neural network training

  • method (string, optional) – Samplers. LINNA supports emcee and `zeus`(default)

  • nbest (int or list of int) – number of points to include in the training set per iteration according to the optimizer

  • chisqcut (float, optional) – cut the training data if there chisq is greater than this value

  • loglikelihoodfunc (callable, optional) – function of model, data , inverse of covariance matrix and return the log liklihood value. If None, then use gaussian likelihood

Returns

MCMC chain 1d array: log probability of MCMC chain

Return type

nd array

linna.main.ml_sampler_core(ntrainArr, nvalArr, nkeepArr, ntimesArr, ntautolArr, meanshiftArr, stdshiftArr, outdir, theory, priors, data, cov, init, pool, nwalkers, device, dolog10index, ypositive, temperatureArr, omegab2cut=None, docuda=False, tsize=1, gpunode=None, nnmodel_in=None, params=None, method='emcee', nbest=None, chisqcut=None, loglikelihoodfunc=None, nsigma=3)[source]

LINNA main function

Parameters

  • ntrainArr (int array) – number of training data per iteration

  • nvalArr (int array) – number of validation data per iteration

  • nkeepArr (int array) – number of autocorrelation time to be kept

  • ntimesArr (int array) – number of autocorrelation time to stop mcmc

  • ntautolArr (float array) – error limit of autocorrelation time

  • meanshiftArr (float array) – limit on mean shift of parameter estimation from the first and second half of the chain

  • stdshiftArr (float array) – limit on std shift of parameter estimation from the first and second half of the chain

  • outdir (string) – output directory

  • theory (function) – theory model

  • str (priors (dict of) – [float, float]): string can be either flat or gauss. If the string is ‘flat’, [a,b] indicates the lower and upper limits of the prior. If the string is ‘gauss’, [a,b] indicates the mean and sigma.

  • data (1d array) – float array, data vector

  • cov (2d array) – float array, covariance matrix

  • init (ndarray) – initial guess of mcmc,

  • pool (object) – mpi4py pool instance

  • nwalkers (int) –

  • device (string) – cpu or gpu

  • dolog10index (int array) – index of parameters to do log10

  • ypositive (bool) – whether the data vector is expected to be all positive

  • temperatureArr (float array) – temperature parameters for each iteration

  • omegab2cut (list of int) – 2 elements containing the lower and upper limits of omegab*h^2

  • docuda (bool) – whether do gpu for evaluation

  • tsize (int, optional) – number of cores for training

  • gpunode (string) – name of gpu node

  • nnmodel_in (string) – instance of neural network model

  • params (dictionary) – dictionary of parameters

  • method (string) – sampling method

  • nbest (int or list of int) – number of points to include in the training set per iteration according to the optimizer

  • chisqcut (float, optional) – cut the training data if there chisq is greater than this value

  • loglikelihoodfunc (callable) – function of model, data , inverse of covariance matrix and return the log liklihood value

  • nsigma (float) – the training point in the first iteration will be generated within nsigma of the gaussian prior

Returns

MCMC chain 1d array: log probability of MCMC chain

Return type

nd array